THE EVOLUTION OF CONSUMPTION: THEORIES AND PRACTICES
ADVANCES IN AUSTRIAN ECONOMICS Series Editor: R. Koppl Associate Editors: J. Birner and M. Wohlgemuth Volume 9:
Cognition and Economics – Edited by E. Krecke, C. Krecke and R. Koppl
Volume 8:
The Dynamics of Intervention: Regulation and Redistribution in the Mixed Economy – Edited by P. Kurrild-Klitgaard
Volume 7:
Evolutionary Psychology and Economic Theory – Edited by R. Koppl
Volume 6: Austrian Economics and Entrepreneurial Studies – Edited by R. Koppl
ADVANCES IN AUSTRIAN ECONOMICS
VOLUME 10
THE EVOLUTION OF CONSUMPTION: THEORIES AND PRACTICES EDITED BY
MARINA BIANCHI Universita` degli studi di Cassino, Italy
Amsterdam – Boston – Heidelberg – London – New York – Oxford Paris – San Diego – San Francisco – Singapore – Sydney – Tokyo JAI Press is an imprint of Elsevier
JAI Press is an imprint of Elsevier Linacre House, Jordan Hill, Oxford OX2 8DP, UK Radarweg 29, PO Box 211, 1000 AE Amsterdam, The Netherlands 525 B Street, Suite 1900, San Diego, CA 92101-4495, USA First edition 2007 Copyright r 2007 Elsevier Ltd. All rights reserved No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without the prior written permission of the publisher Permissions may be sought directly from Elsevier’s Science & Technology Rights Department in Oxford, UK: phone (+44) (0) 1865 843830; fax (+44) (0) 1865 853333; email:
[email protected]. Alternatively you can submit your request online by visiting the Elsevier web site at http://elsevier.com/locate/permissions, and selecting Obtaining permission to use Elsevier material Notice No responsibility is assumed by the publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN: 978-0-7623-1452-2 ISSN: 1529-2134 (Series) For information on all JAI Press publications visit our website at books.elsevier.com Printed and bound in the United Kingdom 07 08 09 10 11 10 9 8 7 6 5 4 3 2 1
CONTENTS LIST OF CONTRIBUTORS
vii
ADVISORY BOARD
ix
INTRODUCTION
xi
PART A: CONSUMPTION AS AN ACTIVITY FROM CARL MENGER’S THEORY OF GOODS TO AN EVOLUTIONARY APPROACH TO CONSUMER BEHAVIOUR Wilhelm Ruprecht
3
WHAT SHALL I DO? (OR WHY CONSUMER THEORY SHOULD FOCUS ON TIME-USE AND ACTIVITIES, RATHER THAN ON COMMODITIES) Ian Steedman
31
IDIOSYNCRATIC LEARNING, CREATIVE CONSUMPTION AND WELL-BEING Marina Di Giacinto and Francesco Ferrante
41
PART B: UNCERTAINTY, NOVELTY, AND CHOICES A SHACKLEAN APPROACH TO THE DEMAND FOR MOVIES John Sedgwick v
77
vi
CONTENTS
THE EVOLUTION OF ENTERTAINMENT CONSUMPTION AND THE EMERGENCE OF CINEMA, 1890–1940 Gerben Bakker CINEMA AND TV: AN EMPIRICAL INVESTIGATION OF ITALIAN CONSUMERS Andrea Sisto and Roberto Zanola
93
139
PART C: SOCIAL COMPETITION AND INTERDEPENDENT PREFERENCES SMOKE SIGNALS: ADOLESCENT SMOKING AND SCHOOL CONTINUATION Philip J. Cook and Rebecca Hutchinson
157
FASHION, GROWTH AND WELFARE: AN EVOLUTIONARY APPROACH Andreas Chai, Peter E. Earl and Jason Potts
187
FASHION: WHY PEOPLE LIKE IT AND THEORISTS DO NOT Luciano Andreozzi and Marina Bianchi
209
DOES CONTEXT MATTER MORE FOR SOME GOODS THAN OTHERS? Robert H. Frank
231
LIST OF CONTRIBUTORS
Luciano Andreozzi
Universita` degli studi di Trento, Italy
Gerben Bakker
London School of Economics and Political Science, UK
Marina Bianchi
Universita` degli studi di Cassino, Italy
Andreas Chai
Max Planck Institute of Economics, Germany
Philip J. Cook
Duke University, Durham, NY, USA
Marina Di Giacinto
Universita` degli studi di Cassino, Italy
Peter E. Earl
University of Queensland, Australia
Francesco Ferrante
Universita` degli studi di Cassino, Italy
Robert H. Frank
Cornell University, Ithaca, NY, USA
Rebecca Hutchinson
US Census Bureau, USA
Jason Potts
University of Queensland, Australia
Wilhelm Ruprecht
GDV European Office, Brussels
John Sedgwick
London Metropolitan University, UK
Andrea Sisto
University of Eastern Piedmont, Italy
Ian Steedman
Manchester Metropolitan University, UK
Roberto Zanola
University of Eastern Piedmont, Italy
vii
This page intentionally left blank
ADVISORY BOARD Don Bellante University of South Florida, USA
Uskali M €aki University of Helsinki, Finland
James Buchanan George Mason University, USA
Ferdinando Meacci Universit‘ a degli Studi di Padova, Italy
Stephan Boehm University of Graz, Austria
Mark Perlman University of Pittsburgh, USA
Peter J. Boettke George Mason University, USA
John Pheby University of Luton, England, UK
Bruce Caldwell University of North Carolina, USA
Warren Samuels Michigan State University, USA
Jacques Garello Universit0 e d’Aix-Marseille, France Roger Garrison Auburn University, USA
Barry Smith State University of New York, USA
Jack High George Mason University, USA
Erich Streissler University of Vienna, Austria
Masazuni Ikemoto Senshu University, Japan
Martti Vihanto Turku University, Finland
Richard N. Langlois The University of Connecticut, USA
Richard Wagner George Mason University, USA
Brian Loasby University of Stirling, Scotland, UK
Lawrence H. White University of Missouri, USA
Ejan Mackaay University of Montreal, Canada
Ulrich Witt Max Planck Institute, Germany ix
This page intentionally left blank
INTRODUCTION 1. INTRODUCTION The theory of consumer choice fills the opening chapters of any microeconomics textbook. Yet, surprisingly, this position of privilege has not translated into a flourishing of economic research that is comparable to what has happened in other branches of economic reasoning. One possible reason for this has to do with the motivational structure of consumer satisfaction, which is subjective and invisible, and therefore hard to reduce to a uniform quantitative measure. The long debate that marked the beginnings of utility theory bears witness to the elusiveness and complexity of concepts such as utility, desires, satisfaction, and well-being.1 The analytical shift from utility to an axiomatic approach to choice provided a way out. By endowing individuals with the ability to rank their alternatives and subsequently to respect their own rankings, preferences could be said to be well-behaved and guarantee certain equilibrium properties. At the same time, inquiry into the nature of preferences or into how preferences are formed and then transformed into choices could simply be dispensed with. Despite the analytical successes of the axiomatic approach, we see today a re-emergence of the neglected motivational dimension of consumer behavior. Thanks largely to recent research in experimental psychology, behavioral economics and neuroeconomics the traditional economic assumption that unobservable preferences can be inferred from observed choices has been scrutinized, tested, and criticized. In choices that span time and in which the past influences the future, individuals seem less able to predict the utility experienced from their choices than standard rationality presumes. On the contrary, they are prone to errors and appear unable to correct them once they are recognized. The result is that preferences can and do diverge from choices.2 As with negative habits and addiction, repeated and just as regularly regretted shortsightedness, and adherence to conformism, there seems to be great ease of entry into some patterns of choice, but no easy way out. Conversely, and more positively, there are also benign habits, displayed in all those activities we pursue for the sheer joy they bring us: music, visual art, literature, sports and games, scientific research and free exchanges of xi
xii
MARINA BIANCHI
ideas and feelings. Such pursuits defy simple utilitarian rationalization; they are their own reward. What is it in these activities that is so rewarding? And why do not consumers simply substitute them for their negative habits, just as good technologies push out bad? These questions have begun to be addressed by a new literature on the determinants of well-being, one that has enlarged the scope of consumers’ choices beyond purely monetary or material rewards. The difficulty of measuring subjective satisfaction may be one reason for the historical veering away from utility and the analysis of individual choice. A second, though related reason may be the difficulty that consumers’ choices often display social interdependencies. While interactions among firms and their strategic rules have long been analyzed and translated into game theoretic models of market dynamics, the consumer continues to be represented as an isolated figure, a lonely chooser. Yet, if a firm in a competitive market misinterprets market signals, behaves inefficiently, or fails to innovate, it is driven from the market. There are no corresponding mechanisms of error detection and correction for consumers. Few economic models investigate how the forces of social competition and coordination work for consumers, and those that do tend to represent consumers as simply prone to herd behavior. At the same time, the interdependent decisional model of the firm cannot simply be imposed on the consumer. The social and market interdependencies among consumers are more complex. They involve different channels of communication, as well as different price-related mechanisms, with a prevalence of hidden or shadow prices. The result is that we still know little of the processes behind the interaction of individual’s preferences. These two difficulties are at the core of this small collection of essays. Let us start from the latter.
2. THE ROLE OF SOCIAL INTERACTION IN COORDINATING THE PASSIONS: PLAYING THE PASSION AGAINST ITSELF The emergence of economic thinking as an autonomous and independent discipline was preceded by a felicitous period of interdisciplinary thought. At center stage was the study of human nature with its conflicting appetites and desires. Political, ethical, economic, and esthetic concerns all contributed to an understanding of how society could work in the presence of
Introduction
xiii
individual passions. By the eighteenth century, as Albert Hirschman correctly notes, religious precepts and moral commands could no longer provide the binding rules that kept a complex society together. Humankind as it was and not as it ought to be was the theoretical shift that allowed for alternative solutions (see Hirschman, 1977). One such was provided by Hobbes: suppressing the passions through the coercive power of the State. An alternative was offered by Mandeville, Smith, and Hume, each of whom viewed the passions not as evil, but as the driving forces that would allow social rules to emerge. As rules emerged, passions continued to exist but in an environment that made them more likely to strengthen social coordination. Such are the rules of the market, and of political and civil liberties, of reciprocation and friendship, of manners and language. Within this context of socially evolved rules it was not reason that was opposed to the passions, but, as Mandeville put it, stronger passions that successfully played against weaker.3 Sympathy and empathy, though also self-love and shame, envy and greed, found new directions in the rules of competition and industry and in the discovery of new sources of well-being. As in a game of skills where each player’s aggressiveness and pride is turned not against the opponent but toward finding unpredictable, surprising ways to win the game, so it was in these early representations of the societal game. It was Hayek’s great merit to have resurrected this tradition of thought and given it the coherence of an analytical framework. To use reason effectively for Hayek meant to recognize its limits in regulating the relations between differently inclined yet still reasonable beings (Hayek, 1967, p. 84). As for the social philosophers before him, so for him too social rules, values, and reason are not deliberately made by man but are the product of social evolution, the unplanned result of planning and mutually interdependent individuals (ibid., p. 86). Consequently, the economic problem for Hayek is how to utilize a knowledge that exists only in the separate and often conflicting beliefs of all members of society (Hayek, 1960, p. 25). The system of market rules here provides the coordinating solution. Market signals become the carriers of the information that allow individuals to judge how best to use the resources over which they have immediate knowledge. Yet, in so doing, they unwittingly serve the needs of other people. (Hayek, 1988, p. 77). Hayek, however, did not explore how the coordinating rules of the market specifically work in the case of different economic agents. We know that if firms mistakenly misinterpret market information, profit losses will soon impair their competitiveness. But how the market unveils consumers’ utility losses is much less clear. Prices alone are a poor signal of future experienced utility and utility losses do not necessarily translate into monetary losses.
xiv
MARINA BIANCHI
Since errors, like motivations, are less observable than in the case of the firm they are also more difficult to correct. Still, cooperation and rivalry works for consumers too, though via the creation of additional signals of communication besides income, prices and profits. The thinkers of the seventeenth and early eighteenth century were more audacious in spelling out these mechanisms. The mercantilists were the first to discover (and turn to advantage) the attraction for consumers of the new goods brought to the market by the expansion of trade. For some this was a cause of concern, for others a sign of progress. Barbon (1690), for example, viewed trade as opening up new spaces of interaction: the cities. By allowing for proximity, and visibility with anonymity, cities smoothed and enlarged the flows of communication and socialization. More importantly, though, through emulation and rivalry, they favored the discovery of new ways of satisfying the wants of the mind, which are unlimited. For Mandeville (1924 [1714]) vanity, self-love, and prodigality, exercised through market exchanges, allowed ingenuity to flourish.4 Smith (1978 [1762-63, 1763-64], pp. 335–337, 448) went so far as to analyze in a general way the pleasureyielding properties of goods, consistently with his insistence that it is consumption, ultimately, that gives point to all production. What we find in these early debates that subsequently has been lost is that social interaction channeled motivations toward new and competing ways of living together. It is through the push of social emulation and competition that consumers act entrepreneurially and enlarge the sets of options from which they can freely choose. In a way Hayek followed in the steps of these eighteenth-century thinkers. His first theoretical contribution was in cognitive psychology (see Hayek, 1952).5 Strangely, however, this involvement with the working of the mind in his case did not translate into a theory of purposeful creative agents in the market. The preconditions are there in minds that learn, introduce changes, and innovate, but they are not explored.6 As Kirzner has put it, Hayek’s knowledge problem has to do with errors that are self-revealing and self-correcting through the working of the market process. Hayek addresses the problems for individuals of having made wrong assumptions on the willingness to pay or sell by other market participants (Kirzner, 1992). But there are other errors that are not selfrevealing and require to be discovered by an alert agent, ready to exploit gainful opportunities.7 Kirzner, however, like Hayek, does not allow that the consumer might be such an entrepreneurial agent. Nevertheless, the tradition on which Hayek built suggests that it is within this framework of trial and error procedures that consumers’ choices and interactions should
xv
Introduction
be analyzed. Emulation, discovery and error-correction are its defining concepts.
3. SATIETY AND BEYOND There are certain consumption activities that seem to escape the law of satiety. The pleasure of listening to music does not diminish with repeated consumption. Gary Becker was the first to address this problem and to reconcile it with the conventional tools of economic choice (Stigler & Becker, 1977). His explanation was economical too. Repeated exposure increases the efficiency of consuming that particular activity and therefore lowers its opportunity cost or shadow price. Preferences remain intact but the use of commodities such as music increases simply because they have become cheaper relatively to other consumption goods. The problem with this ingenious explanation is that the consumption activity it represents is purely fictitious. The ‘‘commodity’’ consumers are supposed to be engaged in is thought of as a continuous stream of an unchanging experience. Instead, the music we continue to listen to without satiation changes with every exposure, either because it reveals new, unexpected, and unnoticed passages and nuances, or because we constantly shift to new pieces (see Bianchi, 2002). If choices reveal preferences here they reveal a love for variety, novelty and surprise.8 In commodities such as music it is this ability to produce change that defeats satiety, engages the mind, and generates increasingly pleasurable feelings. Becker’s insistence on acquired proficiency is obviously crucial, but not without an understanding of the motivational basis of choice. We may become immune to satiation with music, but not, no matter how proficient we become in their use, with hammers, toothpaste, or shoe-polish, though, when changes occur in these, we might be intrigued and increase our consumption. When we adopt this different perspective, of mental rather than purely physical satiety, many things change in both the traditional and the Beckerian theory of consumption choices (see Witt, 2001). First, even if many activities challenge satiety, we do become satiated: a piece of music too often listened to, a genre of literature too much visited, or a research project fully explored, eventually becomes dull and tiresome. Yet, contrary to physical satiety, mental satiation does not correspond to a position of equilibrium and rest. Satiation here implies boredom and boredom produces unrest and displeasure, hence a desire to engage in searches for new options (see Berlyne, 1971).
xvi
MARINA BIANCHI
The active production and experience of goods and activities that can be novel and challenging is the result of this condition of unrest. The consumer is like Kirzner’s entrepreneur who uses his skills of alertness for discovering new profit opportunities. For the consumer too, the discovery of yetunexplored uses and combinatory solutions may be a constant source of both efficiency and utility gains. Yet, looking for activities that become the source of new stimuli may not always yield positive results. There are numerous activities whose enjoyment, as in the case of listening to music, does not decrease with consumption, but that, unlike music, are harmful. Drug use and gambling, but also the use of violence, can be of this sort (see Scitovsky, 1992 [1976]). Again, it has been Becker’s great merit to have first drawn attention and analytical resources to this problem (see Becker, 1996; Becker & Murphy, 1988). With his distinction between positive and negative addictions he shows the differences, but also the similarities between these two forms of consumption in the sense that both are cumulative and rest on complementarities between past and present exposure. Since Becker many studies, including, recently, some coming from neuroscience, have helped clarify the mechanisms of negative addiction. Several explanations have been proposed that address that peculiar conflict of choice that characterizes advanced negative addictions: that users find themselves engaged in unwanted but compulsive consumption. Some works invoke different forms of myopia such as the short-sightedness of piecemeal actions (Herrnstein & Prelec, 1992) or that of magnifying proximate rewards relatively to more distant ones (see Loewenstein, Read, & Baumeister, 2003).9 Other research has started to uncover the neuro-physiology of the brain that is at work in the case of drug use (Bernhein & Rangel, 2004). Exposure to addictive substances sensitizes users to environmental cues that trigger mistaken usage. These occur because drug substances interfere with the learning process that associates environmental cues with the anticipation of pleasure. With repeated use, some cues associated with past consumption – smelling alcohol, seeing a cigarette – cause that part of the brain responsible for hedonic responses to anticipate grossly exaggerated pleasure responses that make consumption impossible to resist.10 The recognition that in many consumption activities satiety is challenged and preferences escalate, causing both positive and negative mistakes, tells us that the relation between motivations and choices is of a complex nature and still in need of being explored. Its analysis has to include both the intrapersonal and temporal dimensions as well as the interpersonal, social dimension of choices. In this process, motivations and preferences have not
Introduction
xvii
only a cognitive dimension but also an affective one, as the economic writers of the eighteenth century particularly recognized. Several converging studies, from different disciplinary points of view, have started to address the affective, emotional dimension of motivations.11 Again, contributions of neuroscience have been especially revealing in stressing the importance of affect in decision making. Between affect and cognition there can be competition and conflict, as when emotions distort cognitive judgments. But also, more surprisingly, collaboration (Camerer, Lowenstein, & Prelec, 2005, p. 29). As many studies of individuals who have suffered prefrontal brain damages have been able to show, the inability of these patients to act sensibly according to their knowledge is due not to any lack of cognitive skills, which are intact, but to their inability to activate an emotion-related signal (Damasio, 2003, pp. 144, 148). Emotions, it seems, are not ancillary to reason but are part of the efficient working of reasoning.12
PART A: CONSUMPTION AS AN ACTIVITY Menger’s Theory of Goods and the Evolution of Consumption In asking how a good becomes a good, i.e. something whose use is valued by the consumer, Menger transforms the consumer into an active agent. The consumer is no longer assumed simply to react to the potential gains and losses associated with the price changes of goods already given; on the contrary, consumers are agents who give a simple thing the status of good through their specific knowledge and needs. Menger’s theory of goods is explained, discussed, and developed at length in Ruprecht’s paper ‘‘From Carl Menger’s Theory of Goods to an Evolutionary Approach to Consumer Behavior’’. The prerequisites for a good to become a good are essentially two: the presence of a human need and the individual knowledge of the good’s ability to answer the need. But how is this knowledge acquired? Menger had anticipated Hayek in trying to explain how social institutions that serve the common welfare come into being without a common will (Menger, 1963 [1883]).13 For Menger the evolution of languages, the rise of new localities, the origin of money, or the development of market relations, are a result of the increased knowledge of one’s own interests that accrues to individuals engaged in formal and informal interchanges (ibid., pp. 154–155).14
xviii
MARINA BIANCHI
Menger, however, does not tell us how this gradual activation of individual interests through social interaction works in the case of consumption practices. Ruprecht in his paper starts to look at complementary models that might provide a direction. He selects those concepts of evolutionary biology and evolutionary psychology that can create an interface between the acquisition of individual knowledge and broader patterns of consumption behavior. The associative learning of primary and secondary reinforcement provides the first step, but it is the cognitive and social learning that are most relevant. The cognitive learning is embedded in the ability to delay consumption through the creation of tools – Menger’s second order goods. Social learning, for its part, allows consumers to take advantage of each other’s experiences when facing unknown situations. In spelling out these processes of learning, however, additional explanatory concepts are needed, such as the physical availability of goods (and its dynamics) and ‘‘tightness’’ – the contextual institutional framework of changing consumption patterns.
The Role of Time If, following Menger, we frame consumption in terms of activities that we want to perform, rather than in terms of quantities of commodities that we want to consume, a whole array of new theoretical and factual perspectives opens up. Consumption becomes a process that involves many stages and, consequently, the constraints become more complex in nature. They involve not only the traditional income and price constraints but knowledge and experience, and, importantly, time. Consumers learn very quickly that consumption activities take time and that knowledge means having to plan actions that have future effects, or that might be conflicting. In his paper ‘‘What Shall I Do?’’ Ian Steedman discusses the analytical consequences of inserting time into consumption activities. Steedman’s direct inspiration in fact comes from the work of Gossen, who was the originator of marginal utility theory, but whose concept of utility, later totally lost, depended entirely on the frequency and duration of the consumption experience rather than on the quantity consumed. According to Steedman, when the time dimension is taken seriously, consumers not only face two constraints, money and time, the second is at least as powerful as the first. Indeed, it is always to the specific time requirements of each activity that one is ultimately forced to bind one’s own preferences. Thus, when income constraints are below satiation – as might often be the case – any change in income or prices will also affect the allocation of time among
xix
Introduction
various activities. If a time-saving activity becomes more expensive and has to be replaced by one that is more time consuming, this will subtract time from other consumption activities that must then be re-planned.15 The consequent effects on demand are difficult to predict. Inferior and Giffen goods will tend to become more likely, as will new forms of complementarity and substitutability. For example, a drop in the prices of books will not cause the demand for books to increase if one does not have the time to read them. Rather, it might increase the demand for CDs that one can listen to while driving. After a while, activities whose time use is more flexible become more attractive, and this strategy might translate into no reading at all. According to Steedman, a theory of choice centered on different uses of time is more amenable to spelling out the distinct natures of the motivations behind consumption activities, as well as the social dimension of consumption, than the usual time-less theory in economic textbooks. The Role of Learning The time dimension of consumption activities is discussed also in the paper of Marina Di Giacinto and Francesco Ferrante, ‘‘Idiosyncratic Learning, Creative Consumption and Well-being’’. Here what is stressed is the time required in the investment phase, when one forms skills and knowledge through exposure and experience. But learning to know is also learning to like or dislike; learning always feeds back on preferences. The model of preference formation they propose tries to avoid both the characteristics of unbounded rationality in Becker’s models – where agents are assumed to be able to map all their inter-temporal preferences and therefore to know the marginal rates of substitution of present versus future consumption – and the myopic rationality of models such as Pollack’s, where preferences depend on the past but cannot anticipate the future. They adopt a more realistic approach where individuals are able to learn their preferences through consumption feedbacks. Such learning, however, is localized and specific and the skills thus gained are ‘‘illiquid’’, not easily transferable among different individuals. The lock-in effects that might derive from this form of learning can be particularly relevant for creative and cultural activities which require, before they can be appreciated in full, the acquisition of consumption skills. Embedded in the way learning proceeds, this form of path-dependency may be inevitable. Still, it can be a cause for regret and of the loss of some of the sources that make life enjoyable and full. The main thrust of the paper is akin to a similar type of argument made by Scitovsky (1992 [1976]). Creative consumption, he argued, requires a
xx
MARINA BIANCHI
more general type of knowledge, one that is able to connect and integrate the separate forms of local information, but also one that is scarce because of increased specialization. A possible solution envisaged by Scitovsky and suggested also in the paper lies in the existence of competing sources of consumption exposure – an enlarged set of consumption experiences – to be coupled with the more critical attitudes that come from a liberal education.
PART B: UNCERTAINTY, NOVELTY AND CHOICES The Demand for Cinema Imagine an industry whose revenue dynamics are almost chaotic, following a path that bifurcates into high revenues and long product lives and low revenue and short product lives. There is no way to impose order on these dynamics and make a preferred outcome prevail since indicators such as costs or quality differences do not seem to affect the outcome. Can such as industry exist? Yes, for these are the characteristics of the film industry, where success is rare and wholly unpredictable but so powerful when it occurs as to obscure the prevalence of failures. The great uncertainty that characterizes the film industry is shared also by other creative industries, by the music and entertainment industry, and by the book, tourism, and fashion industries, where the rate of change and innovation is high and where demand is difficult to manage or predict (De Vany, 2004). John Sedgwick in his ‘‘A Shacklean Approach to the Demand for Movies’’ addresses this peculiar uncertainty from the point of view of film consumption. Films are the types of commodities one must pay for before knowing the full intensity of the enjoyment they will deliver. Even with accumulated experience, the likelihood of consumers mis-predicting their preferences is high. Yet, given the ubiquity and popularity of film consumption, consumers must have developed some heuristics to partially master this uncertainty and, importantly, to transform it into a source of enjoyment. Sedgwick proposes to model this peculiar form of consumption using Shackle’s potential surprise approach to choice. Shackle, a student of Hayek, had made novelty the central focus of individual choices. Choices are originative, creative of new possible options, the result of the imaginative power of the mind. No choice can escape the uncertainty and the potential surprise of enjoying or suffering unrealized expectations (see Shackle, 1972). Following Shackle, Sedgwick identifies the prospective utility gains or losses a filmgoer has learned to associate with specific film markers (actors, genres, directors, friends’
Introduction
xxi
opinions, etc.) and maps these against the degree of potential surprise that s/he is able to anticipate. Within this framework, films may differ in their utility profiles. They may show different levels of potential surprise – of potential mis-predicted outcomes – and, for any such level, they may promise low utility losses but also low utility gains, or, conversely, high disappointments but also great pleasure. Depending on how experimenting and noveltyloving consumers are, they can trade off the safety and lower disappointments of films with low levels of potential surprise, with the greater prospective losses but also the greater excitement of less predictable films. The obvious counterpart of this framework of choices can be found in the way producers act: they make the best of this consumer trade-off by trying to reduce the risks of potential losses without reducing the potential gains.
How Cinema Won over Alternative Forms of Entertainment The incredible early popularity of films and their rapid diffusion, is historically reconstructed in Gerben Bakker’s paper ‘‘The Evolution of Entertainment Consumption and the Emergence of Cinema, 1890–1940’’. His thesis is that it was the strength of consumers’ demand that made possible the development of the cinema industry as a distinct and major form of entertainment. He gives three different arguments for this. In contrast to the accepted view that sees cinema as the invention of few genial men, Bakker shows the existence of a time lag of roughly 12 years between the availability of film technology – the first film projection of the Lumiere brothers in 1895 – and the real take-off of films as an independent form of entertainment in 1905–1907. During this period short films were projected simply as side shows of stage and vaudeville attractions or at fairs. Bakker’s suggestion is that this state of affairs could have continued for a long time if not forever had the demand for entertainment not risen steadily in those years as part of a more general increase in incomes and leisure time. Additionally, as cinema spread, a taste for this completely new form of entertainment started to develop, most strongly in the US, where the demand for films soon replaced that for live entertainment and continued to do so from these first years till the late thirties. Live entertainment, though declining, remained instead a strong component of consumers’ demand in both England and France. Bakker develops a method that allows him to detect what portion of cross-country differences in amount of entertainment consumed was due to a technology/price effect and what to a difference in preferences. He shows that while technological differences could explain much of the consumption
xxii
MARINA BIANCHI
differentials between the US and the UK, in the case of France these technological differences were minimal so that differences in entertainment consumption was mostly a matter of taste. Finally, through a formal comparative growth and simulation analysis that adds to the mix also other recreation products, he shows that, though the consumption of all these products increased rapidly in the period under consideration, audiovisual entertainment consumption grew twice as fast as the average for all other groups. Cinema and TV The demand for cinema is analyzed also in Andrea Sisto and Roberto Zanola’s paper ‘‘Cinema and TV: An Empirical Investigation of Italian Consumers’’. Here they stress that particular aspect of cinema consumption which, like other creative activities, makes it strongly dependent on previous exposure. Past consumption positively influences present consumption. They use Becker and Murphy’s model of positive and negative addiction to test this form of inter-temporal complementarity, but they also extend the model to include a comparison with a second addictive good, TV movie consumption. Their data (aggregate regional monthly time series from January 2000 to December 2002 for the 20 Italian regions) show that both past and future consumption have a positive impact on present consumption and, though less strongly, on future consumption. This last feature can be interpreted as a sensitivity of the demand for cinema not only to prices, but, and more importantly, to the opportunity costs of time. These can be assumed to increase with age and with the increasing availability of more flexible, less time consuming and cheaper entertainment activities. What is interesting among their results is that the demand for TV movies is also positively correlated with the demand for cinema, though there is a difference between weekdays and weekends. The two activities are complementary except at weekends, when TV consumption decreases the demand for cinema.
PART C: SOCIAL COMPETITION AND INTERDEPENDENT PREFERENCES Smoke Signals When consumption is seen in its social dimension, consumption activities show themselves to be also a means of communication. How one consumes conveys
xxiii
Introduction
information about one’s personality, ideas, identity and is influenced in turn by the information received from the activities of other consumers. Philip Cook and Rebecca Hutchinson in their paper, ‘‘Smoke Signals. Adolescent Smoking and School Continuation’’, adopt this perspective to explain how American adolescents might decide to start smoking. Previous analyses, in order to explain addiction to smoking, have invoked subjective tastes, individual discount rates, and, in the case of Becker’s model, the lower opportunity costs determined by the accumulation of past consumption experience. But these explanations all focus on consumers taken in isolation. Smoking, instead, Cook and Hutchinson argue, also has high symbolic connotations that may work as a strong signal in social interaction. Smoking is usually associated with being cool, sophisticated, autonomous, and non-conformist. For adolescents these may be positive connotations. At the same time, for adolescents smoking also sends negative signals, such as being a poor student and of being off track at school. Cook and Hutchinson show that smoking is a strong (negative) predictor of both school graduation and college matriculation for both sexes, while drinking is not. This correlation holds even after having adjusted for family structure and other socioeconomic characteristics, such as race, education levels, and the price of cigarettes. If the decision to smoke among American adolescents revealed only their time preferences – with more present-oriented students more inclined to drop out of school and to smoke – this would be true also for alcohol, though it is not. Additionally, the data show that the percentage gap between smokers and non-smokers who remain in school increases as measured students’ abilities increase. This information seems to reinforce the view that smoking works as a signal. For students who do well at school and have high attitude scores, the negative signal of smoking – that of being identified with failing at school – weighs more heavily than the positive image of being cool. The opposite is true for those who do not perform well at school and have low abilities: for these it is the positive signal of rebelliousness and distinction that becomes more attractive. Being off track then provides an additional incentive to smoke. There remain interesting questions: why is smoking and not alcohol an identity signal among adolescents, and does the observed behavior extend to other countries as well?
Fashion 1. Fashion is another form of social interaction among consumers that has always been troublesome for economists to explain. As a result, it has remained always at the margins of economic thinking. Moreover, it has been
xxiv
MARINA BIANCHI
tagged with the normative judgment that fashion is frivolous, wasteful, fit for the weak, and so on. In answer to the view that stigmatizes fashion as wasteful, the paper by Andreas Chai, Peter Earl and Jason Potts, ‘‘Fashion, Growth and Welfare: An Evolutionary Approach’’, suggests an alternative option. Imagine a world without fashion cycles and the associated coordinated renewals of consumer choice. What would happen? Consumers would be deprived of the opportunity to learn and enlarge their experience and consumption capabilities, and the business of fashionrelated firms would stagnate. In an evolving and changing world consumption is a risky activity. Any new item that enters consumption disrupts the complex set of established complementarities and combinatory solutions that form each consumer’s choices and lifestyle. Income is no more the only constraint; attention and skills too have now to be deployed on things that promise to change the consumer’s knowledge base. Any new chosen activity then is likely to generate errors, and these will be more costly the more durable and non-marginal are the goods or activities chosen. Were consumers just isolated entities, there would be no incentive for them to revise their mistakes and explore new opportunities. The result would be increased specialization and more routinized behavior. Consumers would display some of the characteristics of the old with their small obsessions and defensive behavior. Fashion, instead, by stimulating competition and exchanges of experiences, not only favors experimentation with the new, but also makes it easier to abandon old rules and correct past errors. 2. The idea that the main force driving fashion is the desire to explore and experiment with novelty is also at the core of Luciano Andreozzi and Marina Bianchi’s paper. Curiously, most of the models that try to explain fashion completely miss the role of novelty, preferring to focus only on the assumption that preferences are socially shaped. In these models there is no intrinsic motivation that explains the emergence of fashion, only the socially determined desire to conform or be distinctive. The result is that the patterns that emerge are totally self-referential; nothing seems to anchor fashion to one or another good or activity, so long as these have received the mark of approval by those who count – trend setters or members of the upper classes. By contrast, Andreozzi and Bianchi present a model of fashion in which both the intrinsic pleasure of novelty and the mutual interplay of consumers have roles. If consumers are individually attracted, and stimulated by, novel forms of consumption, they also need each other in order to understand and
xxv
Introduction
experiment with novelty. Two variables are crucial in their model. One is the degree of likelihood that consumers start interacting through imitation and mutual learning. When an old fashion has lost its appeal but the advantages of the new are not yet grasped, uncertainty and the risk of making mistakes are greater. This is the moment in which it is more likely that they will turn to each other. The second variable is the relative velocity with which novelty erodes with repeated and extended consumption, which in turn depends on the multiplicity of potential new uses an activity allows. For very low levels of social interaction and when novelty vanishes very quickly fashion cycles do not emerge and everybody chooses according to their own preferences. Fashions start to emerge as social interaction increases and novelty is more durable. Moreover, cycles last longer the longer novelty maintains its appeal. This model explains why some activities, those that are more visible and consumed socially, lend themselves to fashion more than others, and why some enjoy longer sustained popularity than others. Genres in literature, movies, songs, and art forms, styles in dress and furniture, being more complex, internally variable and novelty-producing, tend to display much longerlasting cycles of fashion than, say, individual books, songs, accessories, or decorations. Finally, the model’s simulations show that for very long-lasting novelty appeal, fashion cycles tend to become weak or vanish altogether. This is the case of the ‘‘classics’’: goods whose appeal never fades.
Positional Competition Adopting a geometrical analogy, fashion is a form of horizontal competition, where contestants vie for the different, the new, and the fun-making. Positional interaction is instead a form of vertical competition, where the stakes are relative standing and rank in the social hierarchy. In his paper, ‘‘Does Context Matter More for Some Goods than Others?’’ Robert Frank explores how positional competition can affect the distribution of individual resources among various activities. Not all consumption activities are equally sensitive to the social context within which their participants interact. Competing for all those goods that visibly mark difference, social standing and relative position, such as high incomes, highly paid jobs, big houses and cars, inevitably subtracts resources from goods that are less context-dependent. Expenditures for safety and health, pension savings, time devoted to leisure and to being with friends, the provision and conservation of public goods, are activities that rank low on the positionality scale and that consequently will suffer.
xxvi
MARINA BIANCHI
Frank maintains that this is not simply an inescapable trade-off, but a conflict that leads to a misallocation of resources with consequent welfare losses. Why? Because competing for position is, for Frank, a no-win game: the winner of today is the loser of tomorrow. It is similar to an arms race where escalation can be stopped only through some form of collective action. Is there any empirical evidence of such conflicts? Frank finds indirect confirmation in some forms of social institutions and in legal norms that have emerged and are often enforced by the state with the purpose of mitigating this very sort of competition. Such are the norms that regulate working hours and encourage leisure, those that ensure respect for standards in the work environment, or obligatory retirement saving schemes and compulsory health insurance. Though in this paper Frank focuses only on the negative sides of social competition, not all consumer interactions, as we have seen, are of this type. Moreover, the forms of social control that Frank describes as a means for curbing positional ‘‘arms’ races’’ embody a risk of their own: if extended too broadly they may also cause a loss of the creativity that characterizes horizontal competition, that which makes consumers alert and where consumption practices enlarge and differentiate the set of opportunities.
CONCLUSIONS Starting with Menger, the Austrian economic tradition has always shifted the focus of attention from the problem of equilibrium to that of social order, to the evolution of norms, institutions and practices that favor social cooperation and coordination. Within this tradition competition and markets are not viewed as states, but as processes in which change and errors occur and efficiency is reached but also easily lost. The real economic problem becomes a problem of knowledge – how it is discovered, how it is transmitted. Consumers’ interactions and choices and actual consumption practices play an important role in these evolving forms of sociality. And it is within this framework, that allows for experimentation and learning, that they should be studied.
NOTES 1. For an analytical reconstruction of this problem in both economics and psychology, see Angner (2006).
Introduction
xxvii
2. Kahneman and Thaler (2006, pp. 221–222) distinguish between decision utility and experienced utility. The first, also called wantability, is inferred from choices and used to explain choices; the second refers to the hedonic dimension of an experience. This paper identifies four areas in which errors of hedonic forecasting occur and have been documented: differences in inter-temporal emotional states, differences in the context of choice and the context of experience, biased evaluations of past experience and mis-predicted adaptation. See Loewenstein and Anger (2003) and Kahneman and Krueger (2006). 3. Hirschman – unnecessarily in my view – contrasted these types of explanations, based on processes that re-directed passions toward more beneficial ends, with invisible hand explanations. However, I am aware that Smith’s ‘‘sympathy’’ and his notion that society is nature’s darling are somewhat removed from Mandeville’s ‘‘cunning’’ management by politicians who have to set stronger passions against weaker in order to make ‘‘jarrings in the main agree’’. 4. On these points, see Bianchi (1993) and Bianchi (2001). 5. Abandoned, it was rescued only later in his life and made into a thorough study of the neuro-physiological structure of mental processes (Hayek, 1952). This volume, with its original attention to cognition, has close links with the way Hayek understands and later explains market processes. In relation to this, and relying on a notion of ‘‘adaptive classifier systems’’, William Butos and Thomas McQuade have studied similarities between minds and markets. Both are classifier systems that embody knowledge and generate new knowledge (Butos & McQuade, 2005). Interestingly, also neurosciences lend support to this view: the map of the brain with its specialized functions resembles the map of society (see Camerer et al., 2005). 6. Birner (1999) refers to how Popper, after having received the manuscript from his friend, criticized it rather forcefully for underplaying the creative aspect of the mind, as revealed, for example, in the argumentative and critical function of language. 7. See on this point Earl (2003). 8. See Bianchi (1998, 1999). 9. Once established, these patterns become difficult to break; for habits impose independent claims on behavior that tend to reinforce them. Both the need to compensate for the loss of pleasure a habit induces and the need to avoid the pain of breaking it provide new incentives that prevent individuals from abandoning habits. 10. This part of the brain is the mesolimbic dopamine system (MDS). Experiments have shown that when subjects are presented with a cue associated with a reward, MDS first fires in response to the delivery of the reward and not with the cue, though later it fires only in response to the cue. Addictive substances act directly on the dopamine system and disrupt its ability to form accurate hedonic forecasts. Two separate processes thus seem to be at work in decision-making, a wanting process that responds to the activation of MDS and a liking process that is linked to hedonic rewards. 11. Terminology here differs. In the field of experimental psychology, for example, Deci and Flaste (1996) and Deci and Ryan (2002) distinguish between intrinsically motivated and extrinsically motivated actions, while Apter (2001) prefers paratelic versus telic states. Cultural anthropologist Csikszentmihalyi (1975) focuses on what
xxviii
MARINA BIANCHI
he calls flow experiences to indicate the features of activities that are self fulfilling (provide self-generated rewards). 12. This was particularly evident in situations involving conflicting feelings toward future outcomes (Damasio, 2003, p. 144). According to this author, the role of emotions – joy, sorrow, anger, fear, happiness, sympathy, shame, guilt – is to ensure the regulatory reactions necessary for survival and well-being. The goal of homeostasis is in fact to provide a better than neutral state of life, what we call wellness or well-being (ibid. p. 35). 13. Menger argued that the answer to the origin and change of social institutions should be found in the individual causal factors shaping them, though he adds that, while the societal motives behind beneficially social institutions are easily recognizable, the individual motives involved are hard to discern. For him this explains why common will types of explanation are the most obvious and the ones most readily invoked (Menger, 1963 [1883], pp. 159, 152). As is discussed in Koppl and Whitman (2004), also L. von Mises and his Austrian circle based their understanding of the social on theory and universal laws but also on the subjective meaning of an action. 14. Loasby is perhaps the economist who has most fully explored the connection between economic growth and the growth of knowledge (see for example Loasby, 1999). 15. For example, moving to the suburbs and having to commute. As early as 1959 Scitovsky, anticipating what came to be called the Baumol or cost disease in the live arts, argued that the productivity lags that inevitably afflict activities such as those devoted to personal services, will cause substitution away from, and will thereby penalize, more creative activities (see Scitovsky & Scitovsky, 1959).
REFERENCES Angner, E. (2006). Is it possible to measure happiness? The measurement-theoretic argument against subjective measures of well-being. Mimeo. University of Alabama. Apter, M. J. (Ed.) (2001). Motivational styles in everyday life. A guide to reversal theory. Washington, DC: The American Psychological Association. Barbon, N. (1690). A discourse of trade. London: Tho. Milbourn. Becker, G. S. (1996). Accounting for tastes. Cambridge, MA.: Harvard University Press. Becker, G. S., & Murphy, K. (1988). A theory of rational addiction. Journal of Political Economy, 96, 675–700. Bernhein, D., & Rangel, A. (2004). Addiction and cue-triggered decision processes. American Economic Review, 94(5), 1558–1590. Berlyne, D. E. (1971). Aesthetics and psychobiology. New York: Appleton Century Crofts. Bianchi, M. (1993). How to learn sociality: True and false solutions to Mandeville’s problem. History of Political Economy, 25(2), 209–240. Bianchi, M. (Ed.) (1998). The active consumer. Novelty and surprise in consumer choice. London: Routledge. Bianchi, M. (1999). Design and efficiency. New capabilities embedded in new products. In: P. Earl & S. Dow (Eds), Knowledge and economic organization, essays in honour of Brian Loasby (Vol. 1, pp. 119–138). Cheltenham: Edward Elgar.
Introduction
xxix
Bianchi, M. (2001). The infinity of human desires and the advantages of trade: Nicholas Barbon and the wants of the mind. In: P. Groenewegen (Ed.), Physicians and political economy. Six studies of the work of doctor economists (pp. 48–66). London: Routledge. Bianchi, M. (2002). Novelty, preferences, and fashion: When goods are unsettling. Journal of Economic Behavior and Organization, 47, 1–18. Birner, J. (1999). Making markets. In: P. Earl & S. Dow (Eds), Knowledge and economic organization, essays in honour of Brian Loasby (Vol. 1, pp. 36–56). Cheltenham: Edward Elgar. Butos, W., & McQuade, T. (2005). The sensory order and other adaptive classifying systems. Journal of Bioeconomics, 7, 335–358. Camerer, C., Loewenstein, G., & Prelec, D. (2005). Neuroeconomics: How neuroscience can inform economics. Journal of Economic Literature, 44, 9–64. Csikszentmihalyi, M. (1975). Beyond boredom and anxiety. San Francisco: Jossey-Bass Publishers. Damasio, A. (2003). Looking for Spinoza. Joy, sorrow, and the feeling brain. Orlando: Harcourt. Deci, E., & Flaste, R. (1996). Why we do what we do: Understanding self-motivation. New York: Penguins Books. Deci, E. L., & Ryan, R. M. (Eds). (2002). Handbook of self-determination research. Rochester, NY: University of Rochester Press. De Vany, A. (2004). Hollywood economics. How extreme uncertainty shapes the film industry. London: Routledge. Earl, P. E. (2003). The entrepreneur as a constructor of connections. In: R. Koppl (Ed.), Austrian economics and entrepreneurial studies: Advances in Austrian economics (Vol. 6, pp. 113–130). Hayek, F. A. (1952). The sensory order. An inquiry into the foundations of theoretical psychology. Chicago: Chicago University Press. Hayek, F. A. (1960). The constitution of liberty. London: Routledge. Hayek, F. A. (1967). Studies in philosophy, politics and economics. London: Routledge. Hayek, F. A. (1988). The fatal conceit: The errors of socialism. London: Routledge. Herrnstein, R., & Prelec, D. (1992). Melioration. In: G. Loewenstein & J. Elster (Eds), Choice over time (pp. 235–264). New York: Russell Sage Foundation. Hirschman, A. O. (1977). The passions and the interests. Political arguments for capitalism before its triumph. Princeton: Princeton University Press. Kahneman, D., & Krueger, A. B. (2006). Developments in the measurement of subjective wellbeing. Journal of Economic Perspectives, 20(1), 3–24. Kahneman, D., & Thaler, R. H. (2006). Anomalies. Utility maximization and experienced utility. Journal of Economic Perspectives, 20(1), 221–234. Kirzner, I. M. (1992). The meaning of the market process. Essays in the development of modern Austrian economics. London: Routledge. Koppl, R., & Whitman, D. G. (2004). Rational-choice hermeneutics. Journal of Economic Behavior and Organization, 55, 295–317. Loewenstein, G., & Angner, E. (2003). Predicting and indulging changing preferences. In: G. Loewenstein & E. Angner (Eds), Time and decision (pp. 351–391). New York: Russell Sage. Loewenstein, G., Read, D., & Baumeister, R. (Eds). (2003). Time and decision: Economic and psychological perspectives on intertemporal choice. New York: Russell Sage. Loasby, B. (1999). Knowledge, institutions, and evolution in economics. London: Routledge.
xxx
MARINA BIANCHI
Mandeville, B. (1924 [1714]). The fable of the bees: or, private vices, public benefits. In: F.B. Kaye (Ed.), The fable of the bees (Vol. 2). Oxford: Clarendon Press. Menger, C. (1963 [1883]). Problems of economics and sociology. [Untersuchungen u¨ber die Methode der Socialwissenschaften und der Politischen Oekonomie insbesondere.] Edited, with introduction by L. Schneider. Translated by F. J. Nock. Urbana: University of Illinois Press. Scitovsky, T. (1992 [1976]). The joyless economy: The psychology of human satisfaction (revised edition). Oxford: Oxford University Press. Scitovsky, T., & Scitovsky, A. (1959). What price economic progress. The Yale Review, XLIX, 95–110. Shackle, G. L. S. (1972). Epistemics and economics. A critique of economic doctrines. Cambridge: Cambridge University Press. Smith, A. (1978 [1762-63, 1763-64]). In: R. L. Meek, D. D. Raphael, & P. G. Stein (Eds), Lectures on jurisprudence (Glasgow edition of the works and correspondence of Adam Smith, Vol. 5, pp. 335–337, 488). Oxford: Clarendon Press. Stigler, G. J., & Becker, G. S. (1977). De Gustibus non est disputandum. American Economic Review, 67(2), 76–90. Witt, U. (Ed.) (2001). Escaping satiation. The demand side of economic growth. Berlin: Springer.
Marina Bianchi
PART A: CONSUMPTION AS AN ACTIVITY
This page intentionally left blank
FROM CARL MENGER’S THEORY OF GOODS TO AN EVOLUTIONARY APPROACH TO CONSUMER BEHAVIOUR Wilhelm Ruprecht ABSTRACT A characteristic feature of economic development is the ever-changing structure of consumption patterns. Reducing the explanation of this phenomenon to changing prices, ultimately caused by changes in the availability of goods (or characteristics), would neglect a major force driving this change, namely, the variation of consumer wants and consumer knowledge. The present paper sketches an evolutionary framework for the analysis of consumer behaviour that takes account of these features. For this purpose, Carl Menger’s theory of goods is taken as starting point. Whereas economists after the ‘marginal revolution’ were almost exclusively concerned with the determinants of exchange value and developing price theory, Menger puts as much emphasis on user value as on exchange value. Focusing on how user value changes establishes a connection between Menger’s 19th-century theory of goods and 20thcentury learning theories. The problem of how to get from individual learning processes to aggregate consumption patterns is approached by recollecting the genetic underpinnings of human learning and its The Evolution of Consumption: Theories and Practices Advances in Austrian Economics, Volume 10, 3–29 Copyright r 2007 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 1529-2134/doi:10.1016/S1529-2134(07)10001-6
3
4
WILHELM RUPRECHT
dependence on certain physical and social conditions. Taking into account that these conditions are also dynamic, we are able to interpret collective learning processes as historical events, which renders them useable for the analysis of economic change.
y The subjectivist research program of the Austrian tradition need not be associated at all with an antinaturalistic methodology, but can well be pursued within a causal explanatory framework. (Victor Vanberg)
1. INTRODUCTION A characteristic feature of economic development is the ever-changing structure of consumption patterns. This change is well documented in empirical studies on changing household expenditure patterns (see e.g. Hildenbrand, 1994). These studies usually divide consumption expenditure into various categories such as food, clothing, entertainment and so on, and show how the income share spent on each category alters in time. That there is also change going on within these individual categories, however, is sometimes overlooked. Nonetheless, in the literature on consumer price indices it is recognized that there is continual change in consumption opportunities occurring: new items are introduced and others disappear from the market (see e.g. Nordhaus, 1998). Reducing the explanation of this phenomenon to changing prices, ultimately due to changes in the availability of goods or characteristics (Lancaster, 1966a, 1966b, 1971), would neglect a major force driving this change, namely, the variation of consumer wants and consumer knowledge. To date, economic theory in general and consumer theory in particular have not taken adequate account of these causes of change. A major reason for this lack is axiomatic preference theory, and in particular the completeness axiom. In contrast to the dominant price theoretical approaches, which are concerned with the determinants of exchange value, an evolutionary theory of consumption would be interested in how user value comes into being and how changes in user value take place. For help in rethinking changing consumption patterns from an evolutionary perspective, looking into other disciplines that also deal with consumption behaviour, such as psychology and biology, seems to be a promising strategy. Since each discipline has developed its own specific
An Evolutionary Approach to Consumer Behaviour
5
methods and theoretical perspective on this subject, a precondition for fruitful interdisciplinary exchange is the existence of conceptual interfaces and the development of a ‘common language’. The goal of this paper is to provide such a conceptual interface. For this purpose, Carl Menger’s (1950) theory of goods, which is outlined in the first two chapters of his Principles, is taken as a starting point. Menger’s theory marks a turning point in the history of the utility theory; whereas, following the ‘marginal revolution’, economists were almost exclusively concerned with exchange value and with developing price theory, Menger himself puts as much emphasis on user value as on exchange value. This is indicated in his definition of how a thing is to become a good, which takes account of both features. In Section 2, I scrutinize whether the conditions Menger claims to be necessary for a thing to become a good are really necessary. Then, I propose slight modifications of Menger’s approach so as to allow me to take account of the insights of psychological learning theories which were established more than fifty years after his Principles was published (Section 3). The result is an evolutionary approach to consumer behaviour which allows to explain the ever-changing structure of goods and services on a macro-level.
2. MENGER’S THEORY OF GOODS: CRITICAL REVIEW AND QUALIFICATION In his ‘theory of goods’, outlined in the first two chapters of his Principles, Menger specifies: If a thing is to become a good, or in other words, if it is to acquire goods-character, all four of the following prerequisites must be simultaneously present: 1. A human need. 2. Such properties as render the thing capable of being brought into a causal connection with the satisfaction of this need. 3. Human knowledge of this causal connection. 4. Command of the thing sufficient to direct it to the satisfaction of the need. (Menger, 1950, p. 52)
Two aspects of this definition deserve emphasis. First, Menger’s concept contains scarce goods, which are objects of market exchange, as well as free goods. Second, the Mengerian good is defined in terms of necessary
6
WILHELM RUPRECHT
conditions.1 Menger claims that all four elements of his definition are necessary conditions, i.e. when only one element is lacking, a thing loses its ‘goods’ character. Keeping in mind that definitions can never be wrong, but only more or less useful, I will examine each condition in turn. This prepares the ground for an analysis of how consumption patterns evolve. 2.1. Condition 1: Needs It is intuitively plausible that the existence of a human need is a necessary but insufficient condition for a thing to become a good. Without the possibility of need satisfaction, a thing simply remains a thing. It is possible, of course, that one consumption item satisfies more than one need. For instance, a car can be regarded as a prestige object and, for example, as a source of excitement or cognitive arousal due to its speed. 2.2. Conditions 2 and 3: Objective and Subjective Cause–Effect Relationships For Menger, the existence of an objective cause–effect relationship, which is how I interpret his second condition, is necessary for a thing to become a good. In his third condition, Menger admits that objectively true knowledge about cause–effect relationships cannot be taken for granted. Thus the second and the third conditions hang together and should be discussed simultaneously. In particular, I place their ontological and their epistemological implications at the centre of the analysis. 2.2.1. The Ontological Implications Subjectivity plays a crucial role in understanding how a thing is to become a good. A direct implication of Menger’s definition of goods is that goods are ‘concepts’ which, in contrast to things, do not exist independently of the individuals who conceive of them as goods. Menger (1950, p. 58) was clearly aware of the ontological implications of his third condition.2 Since Menger considers both the objectivity of the cause–effect relationship (condition 2) and the individual consumer’s awareness of this relationship (condition 3) to be necessary conditions, it follows if an objective causal connection between a thing and consumer needs is not known, that thing cannot become a good to a consumer. Taken together, elements two and three of the Mengerian definition allow for the application of a ‘bimodal
An Evolutionary Approach to Consumer Behaviour
7
ontology’3: one implying that objective cause–effect relationships (world) and the subjective perception of them (mind) can change independently of one another.4 Given such a bimodal ontology, not only producers of goods are involved in innovative activities. Even when nothing else in the world changes, changes in the perception of behavioural opportunities can occur. This changing perception can be a major force for socioeconomic change (Witt, 1989a, p. 96). It is not necessary that subjective ideas about cause–effect relations precede their objective realization. It is equally consistent with a bimodal ontology that it be the other way round, i.e. that cause–effect relationships objectively exist but are not recognized or understood immediately: cause– effect relationships may exist without anybody recognizing them. I shall illustrate this latter possibility with two examples: The first is the detection of the impact of citrus fruits on scurvy prevention. Although citrus fruits had, of course, been known for a long time and already had the status of goods because of their caloric content, their juiciness and their taste, their connection with scurvy prevention was an objective novelty. The knowledge about the nutritional causes of scurvy did not emerge until the mid-17th century (Mokyr, 1998b, p. 129). The second example, asbestos, shows that there are also cases in which new knowledge turns former ‘goods’ into ‘bads’ (Witt, 1996a). Because of its resistance to fire, asbestos was once used as a construction material satisfying the need for shelter and safety. Later, it was discovered that asbestos fibres support the emergence of a certain type of cancer. The detection of this detrimental causal connection was certainly a novelty. It had not been detected earlier since there was a time lag between identifying this particular causal effect, in a context where there was multicausality present.5 Briefly, it is obvious that in both instances the positive (citrus fruit – scurvy prevention) or negative (asbestos – cancer) relation existed independently of human knowledge of it. The perception of citrus fruit as good and asbestos as bad, however, could only come with that knowledge. The added value in Menger’s approach consists in his having taken account of change in the conceptual character of an item due to a change in consumers’ perceptions. 2.2.2. The Epistemological Implications The epistemological problem involved in the acquisition of knowledge of cause–effect relations – Menger’s third condition – is next on our list.6
8
WILHELM RUPRECHT
Some fifty years ago, the Peruvian public health service started a campaign to persuade villagers from Los Molinas to boil their drinking water.7 Boiling water can be considered a hygienic innovation that influences life expectancy. For several reasons the causal relationship between drinking unboiled water and diseases is difficult to understand by simple logical induction. It is not water but invisible bacteria which cause the disease. Not every sip of unboiled water necessarily contains sickness-inducing bacteria. Even when there are bacteria within the drinking water, a healthy organism with a functioning immune system may not get ill. The health campaign failed: only 11 out of 200 housewives started boiling the water after a two-year campaign. One reason for this diffusion failure lies in the local belief system, which challenged and prevented the adoption of the boiling practice. According to local belief, boiled water was linked to ‘illness’. The belief system assigned an inherent ‘temperature’ to all foods, liquids and other edible substances that were different from their actual temperature. Only the ill and weak used cooked or ‘hot’ water. If a person became ill, it was inappropriate, for example, to eat pork (very cold) or to drink brandy (very hot). Such extremes of cold and hot had to be avoided by the sick. Cold water, by the same reasoning, had to be heated for the sick, but only for them. The belief system of Los Molinos included no concept of germ contamination of natural water. The sense in boiling water, therefore, lies only in eliminating its coldness, not the bacteria. Because of this association between heated water and weakness and health problems, the inhabitants of Los Molinos learned to dislike boiled water from their early childhood. This illustration serves to establish that, due to the knowledge problem involved in Menger’s third condition, an objective causal connection cannot be considered to be a sufficient condition for ‘goodness’ in a thing. However, I have not asked whether an objective cause–effect relationship between things and need satisfaction has to exist at all. When we switch from ontology to epistemology, we challenge Menger’s (1950) view according to which the objective character of this relationship is a necessary condition for a thing to become a good. Menger himself may have been aware of certain ambiguities of his definition in this respect since he emphasizes a class of items which lack even an indirect causal relationship to need satisfaction. Thus he refers to aphrodisiacs or love potions and amulets which could be subsumed under the label ‘placebos’, very much like the hot–cold-distinction
An Evolutionary Approach to Consumer Behaviour
9
system in Los Molinos. For these items, no objective causal relation to the desired effect that fulfils scientific standards has been established, at least not so far.8 Withholding them from full status as goods, Menger accordingly calls such things ‘imagined goods’. He speculates that their importance will disappear during the process of economic change: ‘‘As a people attains higher levels of civilization, and as men penetrate more deeply into the true constitution of things and of their own nature, the number of true goods becomes constantly larger, and as can easily be understood, the number of imaginary goods becomes progressively smaller. It is not unimportant evidence of the connection between accurate knowledge and human welfare that the number of so-called imaginary goods is shown by experience to be usually greatest among peoples who are poorest in true goods’’ (Menger, 1950, pp. 53–54). On the objectivity requirements for knowledge in the present context, Hayek holds a view that is different from Menger’s. According to him, subjective beliefs matter ‘‘y wherever we have to explain human behaviour toward things; these things must then not be defined in terms of what we might find out about them by the objective methods of science, but in terms of what the person acting thinks about them. A medicine or a cosmetic, for example, for the purposes of social study, is not what cures an ailment or improves a person’s looks, but what people think will have that effect. Any knowledge which we may happen to possess about the true nature of the material thing, but which the people whose action we want to explain do not possess, is as little relevant to the explanation of their actions as our private disbelief in the efficacy of a magic charm is to understanding the behaviour of the savage who believes in it’’ (Hayek, 1979, p. 51). From an ontological point of view, Hayek is certainly right. Menger’s statement could be regarded as an imperialistic way of judging cultures, being not only highly politically incorrect but, even worse, implying an interpersonal and intercultural utility comparison. Bearing in mind that all human beings, Menger included, are socialized within a culture-specific belief system, who has the right to declare certain items that are appreciated by other people to be ‘imagined goods’? The problem, however, does not consist in declaring items to be ‘imagined goods’. It consists rather in assuming some things to be ‘true goods’. Obviously, in this case one has to assume the possibility of objective knowledge about the nature and constitution of things. Indeed, Menger’s propagation of an objective cause–effect relation as a necessary condition could be interpreted as the injection into the analysis of an external scientific
10
WILHELM RUPRECHT
observer who is gifted with objective insight. Nowadays, Popper’s critical rationalism, which emphasized the limitations of scientific knowledge, is a broadly accepted position in epistemology. In contrast to Menger, critical rationalism denies the possibility of knowing the ‘true constitution’ or ‘nature’ of things. More modestly, only the state of non-falsification of certain hypotheses gives these hypotheses a (provisional) state of acceptance in science. Critical rationalism is an example of what Hayek (1979) calls the ‘objective methods of science’. It is incompatible with the idea of ‘true knowledge’. Within the framework of critical rationalism there is nothing like true knowledge; there is only knowledge that is not (yet) falsified. From this point of view, both patient and doctor hold beliefs or hypotheses on cause–effect relations that are not necessarily ‘true’. Does critical rationalism and its rejection of true knowledge make the placebo a paradigmatic case? Some scholars are in favour of this position, the more so as it cannot be denied that these items are traded in markets and, therefore, at least are commodities. The history of medical science reveals that humans have always shown a capacity to set up, to test and to falsify hypotheses about causes and effects. Ex post, many alleged cause–effect relations have been found to be wrong. Viewed from the other side, it is understandable that placebos have had a huge economic relevance: ‘‘The long-term history of medical treatment has been characterized as largely the history of the placebo effect’’ (Gru¨nbaum, 1984, p. 131). The discussion of the placebo case and Menger’s ‘imagined goods’ has implications for Menger’s conditions. It has already been shown that, for ontological reasons, an objectively true cause–effect relationship is not sufficient for a thing to become a good. This is in accordance with Menger’s own position. Now, the impossibility of objective knowledge and the possibility of the placebo case show, in addition, that for epistemological reasons, it cannot be a necessary condition either. The epistemological position of critical rationalism, however, does not imply that a cause–effect relation between a thing and the satisfaction of a human need does not exist. This would be not an epistemological but an ontological inference. The possible existence of objective cause–effect relations is not denied.9 For the question whether a thing is to become a good, however, the subjective sphere simply seems to be the decisive one. Even when a conjectured cause–effect relationship has been falsified, it cannot be concluded that the corresponding items are not goods. For the inhabitants of Los Molinos, unboiled water has not lost its desirable quality for persons
An Evolutionary Approach to Consumer Behaviour
11
in normal health. Boiled water, in spite of its statistically significant positive correlation to health, is not a good to them. When the possibility of objectively true knowledge is denied, an alternative criterion is needed that allows consumers to differentiate between competing cause–effect hypotheses. ‘Non-falsification’ may be a candidate for such a criterion. However, there are cases in which consumers have to choose between several hypotheses that are not (yet) falsified but where they have no opportunity or capability of testing these hypotheses. Cause–effect relations are usually not directly observable. Moreover, there are cases in which huge time lags may separate causes from effects, as in the asbestos example. The possibility of multicausality can further complicate induction for consumers. When people lack the opportunity or the capability of testing competing hypotheses about cause–effect relations, they must often, nevertheless, make up their minds which hypothesis to trust and whether they should act upon it or not.10 Whereas Popper’s critical rationalism is a normative concept, we need a positive approach to how consumers solve their knowledge problems. Mokyr (1999a, p. 4) introduces in this context the concept of ‘tightness’. He defines tightness of knowledge as ‘‘y the degree of confidence that individuals have in the truth of this knowledge and their willingness to act upon it y’’. Given this confidence, consumers transform knowledge into what Mokyr calls ‘recipes’. He characterizes recipes as follows: ‘‘In the consumption process, households do not just purchase consumer goods but convert them into final uses by using a set of techniques I will call recipes y Recipes should not be confused with technologies that are used by the household but generated outside it’’ (Mokyr, 1998a, p. 2). Tightness is a matter of social conventions (Mokyr, 1999b, p. 4). The belief system in Los Molinos can be interpreted in this light as a social convention that determines how people cope with new beliefs and which beliefs diffuse and which do not.11 It is important to note that introducing the tightness concept into consumption theory contradicts a radical ‘tabula rasa’ view, according to which consumers can be manipulated in an arbitrary way.12 Having seen that tightness of knowledge might be a precondition for diffusion, now, the question arises: can ‘tightness’ of beliefs replace the ‘objectivity of cause–effect relations’ as a necessary condition in Menger’s catalogue? As a social convention, tightness can neither be a property inherent in a particular piece of knowledge nor a property inherent in goods. Giving things a good’s property is the act of an individual. An individual may well consider a thing to be a good although the contemporary social conventions
12
WILHELM RUPRECHT
are not at all in favour of or are even against this idea. Since consumers (a) can learn from their own experiences without any social process and any conventions being involved and (b) can choose to hold untight beliefs, ‘tightness’ cannot have the status of a necessary condition.13
2.3. Condition 4: Personal Command Menger’s fourth condition – the ‘‘command of the thing sufficient to direct it to the satisfaction of the need y’’ – requires some clarification. At first sight, one can think of ‘income’ as a necessary precondition for getting ownership or ‘personal command’ of something. Income is subject to systematic changes over time, particularly driven by productivity gains. In his famous essay on The social limits to growth Fred Hirsch (1976) points out, however, that income – at least in absolute terms – is not sufficient for ‘personal command’. This is because the speed and potential for productivity increases differ across the individual sectors of an economy. Hence, increasing income need not necessarily be accompanied by quantitative increases in the physical availability of all kinds of goods. Hirsch stresses the possibility that absolute or social scarcity imposes limits on the availability of certain items he calls ‘positional goods’.14 When income rises on average, through the productivity gains in non-positional sectors, not all consumers will gain command of these ‘positional goods’. Hence, it is not true that economic growth and growing average income automatically provide personal command of everything. The relative income position rather than the absolute income level seems to determine who gets command of the limited amount of such goods as are not sufficiently available to meet all needs.15 However, income increases are not only insufficient for getting command of a thing, they are not necessary either. This becomes obvious when the case of goods that are not traded in markets is considered. The command of free goods presumes their sufficient physical availability but does not require any income at all.16 Having clarified the relationship between income and personal command, we can now address the question of whether personal command is, after all, necessary for a thing to become a good. The example of positional goods shows that wants can be shaped independently of the personal command of a thing. For example, we can easily imagine that there are potential Rolls Royce drivers who would immediately switch to being actual ones given enough income. Apparently, for them, the Rolls Royce is already a good although they cannot afford it.17 This is
An Evolutionary Approach to Consumer Behaviour
13
because consumers learn wants not only by their own consumption experiences but also by observing the consumption of other people. There is a lot of empirical evidence for ‘emulative consumption’ behaviour (see e.g. Bourdieu, 1984 or Rogers, 1995). This is sometimes characterized as keeping up with the Joneses. What is important in the present context is that social imitation provides a strong argument against giving ‘personal command’ the status of a necessary condition for a thing to become a good. We may speculate that Menger included ‘personal command’ as a necessary condition for ‘goods’-ness because he needed this element later in his Principles when discussing the determinants of exchange value.18 From the discussion of whether Menger’s conditions are really necessary for user value to come into being, a qualified concept of goods can be derived. If a thing is to acquire goods-character, the two following prerequisites must hold simultaneously. There must exist 1. A human need for the thing and 2. Human belief on a causal connection between the thing and the satisfaction of this need. Together, these conditions are necessary and sufficient for defining goods and for distinguishing them from related concepts such as ‘things’, ‘artefacts’ or ‘commodities’.19
3. FROM THE ‘THEORY OF GOODS’ TO AN EVOLUTIONARY THEORY OF CONSUMPTION Menger’s question about how a thing is to become a good is certainly evolutionary in character. Its merit is to emphasize the dissemination issue, i.e. the question of how new subjective cause–effect relations are established. So far, evolutionary economics has devoted much more attention to the technological innovation issue, i.e. the question of how new objective cause– effect relations are established. However, unless a vicarious entrepreneur is assumed who knows exactly what consumers want, the demand side processes have to be analysed explicitly.20 It is true that Menger’s theory of goods is open to dynamics, but it is not very concrete about the actual processes. In order to overcome this shortcoming the way ‘men classify external stimuli’21 is to be looked at more closely. The scientific analysis of quality and quality change involves investigating the human mind rather than investigating the physical world.
14
WILHELM RUPRECHT
Fortunately, economists can take advantage of an exchange with the discipline which has specialized in dealing with processes occurring in the human mind, psychology. In this section, therefore, I introduce psychological approaches that provide answers to the positive question of how consumers link causes to effects. In the framework of an evolutionary analysis of economic change, it may not be surprising that psychology also is approached from an evolutionary perspective. Such an approach has been proposed by Witt (2003).22 The basic idea behind what he calls the ‘continuity hypothesis’ is to consider cultural evolution as being based on biological evolution. It is important to note that, unlike in sociobiology, behaviour is not assumed to be genetically fixed. It is, rather, explicitly considered to be open to learning processes. Only the structure of these learning processes is presumed to be hardwired, while their result is not pre-determined. According to evolutionary psychologists, humans have evolved several mechanisms that are capable of linking causes to effects.23 In the present investigation, two different mechanisms – non-cognitive reinforcement learning and social-cognitive learning – are examined more closely.24 The fundamental reservation of some Austrian scholars against the application of a Stimulus–Response scheme, thereby, is invalidated: By integrating a cognitive learning mechanism in the analysis, the intentional and teleological character of human behaviour is sufficiently incorporated. In discussing the necessary conditions of Menger’s definition of a good, two types of learning have already been distinguished: ‘social learning’ and ‘learning from own experiences’. It has been argued that learning by own experiences requires personal command, and that beliefs about cause–effect relationships require ‘tightness’ in order to be transformed into actual behaviour. This simple idea, I suggest, is helpful for identifying the elements that an evolutionary theory of consumption ought to comprise. It is, of course, quite reasonable to expect that the necessary conditions for a thing to become a good, i.e. needs and knowledge, are elements of such a theory. The question of whether a thing is to become a good is, however, an issue that should not be confused with a theory of how consumption patterns change. First of all, actual consumption is possible only when a consumer has command of a thing. When the evolution of consumption patterns is discussed, however, the forces influencing personal command cannot be neglected. Otherwise, the evolutionary theory of consumption and demand would abandon any claim to being an explanation for observable economic phenomena. A second major problem is that Menger’s (1950) ‘theory of goods’ refers to the individual level just as learning of cause–effect relations
15
An Evolutionary Approach to Consumer Behaviour
occurs at the individual level. An evolutionary theory of consumption, in contrast, is interested in phenomena that occur at a more aggregate level. How to cope, then, with this aggregation problem? The fact that ‘tightness’ and ‘personal command’ were not found to be necessary conditions for a thing to become a good, by no means implies that they are irrelevant in an evolutionary explanation of changing consumption patterns. On the contrary, when ‘tightness’ and ‘personal command’ are considered to be preconditions for learning to take place, it is obvious that they should be included among the possible elements of an evolutionary theory of consumption. Their particular advantage is that they can be located at the aggregate level. The connection of ‘tightness’ to the aggregate level is self evident, for ‘tightness’ relies on norms which are inter-individual entities. In the case of personal command, a close connection to the aggregate level is provided by ‘physical availability’ – identified in subsection 2.3 as a necessary condition for personal command to become possible. The physical availability of an item is quantifiable and can be aggregated. It is for these reasons that I propose complementing the two necessary conditions for a thing to become a good, needs and consumer knowledge, by ‘tightness’ and ‘physical availability’ as the third and the fourth elements of an evolutionary theory of consumption (see Table 1).
Table 1. Comparison between Menger’s Conditions for a Thing to Become a Good and the Elements of an Evolutionary Theory of Consumption and their Dynamics. Conditions for a Thing to Become a Good (Menger)
Elements of an Evolutionary Consumption Theory and their Dynamics Elements
Need
Need, reinforcer
Objective cause– effect relation Knowledge of this cause–effect relation
Tightness
Personal command of a thing
Knowledge/beliefs on a cause–effect relation between a thing and need satisfaction Physical availability
Dynamics of Elements Emergence of secondary reinforcers by reinforcement learning Evolution of institutions and norms changes ‘tightness’ Generation and diffusion of new knowledge/ beliefs on a cause– effect relation between a thing and need satisfaction Process innovations and technical progress increase the physical availability of things
16
WILHELM RUPRECHT
Unlike psychology, which usually does not consider the variability of the conditions under which learning takes place, an (evolutionary) economic perspective requires explicitly addressing the dynamics of these conditions. The basic hypothesis underlying this section, therefore, is that all four of these elements are dynamic. In discussing psychological learning theories, subsequently, the processes and forces driving the dynamics of the four elements will therefore be addressed. In subsection 3.1 on ‘reinforcement learning’ it is shown how the dynamics of needs – element 1 – is related to the dynamics of physical availability – element 4. In subsection 3.2 on ‘(social-) cognitive learning’ it is shown how the dynamics of knowledge – element 3 – is related to the dynamics of tightness – element 2. 3.1. Reinforcement Learning Menger himself does not have much to say about needs, their material content and their evolution. When he wrote his Principles, biology and psychology were less developed disciplines than today. Since the behaviouristic school in psychology was established fifty years after Menger’s book appeared, he could not make a link between his theory of goods and reinforcement learning theories. It is important to recognize, however, that both Menger’s concept of a good and its qualified version are compatible with behaviourism. 3.1.1. The Dynamics of Needs (Element 1) Thanks to evolutionary biology, explaining the emergence and evolution of needs or wants has become possible. It is plausible to assume certain universal wants which have been formed during human phylogenesis.25 Humans share with other mammals these wants for certain objects such as air, liquids, nutrients, sleep, warmth, nutrition, sexual activity and maternal care, as has been found in and confirmed by thousands of psychological experiments (see Millenson, 1979). This list can be refined, e.g. in the case of nutrients, by the preference for ‘sweetness’ involving adaptive advantages by allowing mammals to identify quickly food items covering carbohydrates rich in nutrients (see Ruprecht, 2005). It is now argued, in addition, that the ontogenesis of wants can be based on their phylogenesis. According to theories of operant26 and classical conditioning, the learning process starts from reinforcers such as food items, aqueous solutions and the other items listed above. Their reinforcing quality is assumed to have a genetic base. In classical conditioning processes, such
An Evolutionary Approach to Consumer Behaviour
17
stimuli act as primary reinforcers. When they are regularly paired with other items, the latter ones obtain reinforcing potential in their own right and become secondary reinforcers (see e.g. Skinner, 1966; Pulliam & Dunford, 1980). Reinforcement learning may even take place at the level of a culture.27 Both primary and secondary reinforcers can be regarded as goods. The crucial difference between the two types of reinforcers is that the link between cause and effect in the case of primary reinforcers is genetic, whereas in the case of secondary reinforcers it is learned. In further processes of operant and classical conditioning, the secondary reinforcers act as rewarding experiences, and preferences28 of higher order emerge. During the individual learning history, a hierarchy of wants or a ‘preference order’ evolves (Witt, 1987, pp. 112–123).29 Whether reinforcement learning is consistent with Menger’s third condition, i.e. that there must be consumer knowledge about the causal connection between a consumption item and need satisfaction, is questionable. Interpreting reinforcement as cause–effect learning is certainly wrong because in reinforcement learning no knowledge about the world is involved that is independent of the particular behaviour. It can be argued, however, that in some sense, a cause–effect relationship is established. Baker, Murphy, and Valee-Tourangeau (1996, p. 1) characterize the result of reinforcement processes as ‘‘y being able to react appropriately to causes y’’, as opposed to ‘‘y being able to understand them’’.30 3.1.2. The Dynamics of Physical Availability (Element 4) Reinforcement theory is explicit about the conditions under which learning takes place. These conditions involve temporal and spatial contiguity between the conditioned stimulus and the unconditioned stimuli (see Baker et al., 1996). When goods are considered to be multidimensional entities that include several stimuli, this condition is always fulfilled. An additional condition is regularity in the pairing between primary and secondary reinforcers (Baker et al., 1996). In this context, one should remember that under experimental conditions in psychology, the availability of reinforcers is given and controlled. In this respect, psychological reinforcement theory is inherently static. Outside of the laboratories, however, the availability of primary and secondary reinforcers cannot be automatically taken for granted. The physical availability (and number) of reinforcers can vary over time, as has already been mentioned above (see Table 1). In the continuing process of technological change, for example, some reinforcers become more, others become less
18
WILHELM RUPRECHT
available and new items may even become available for the first time. Apparently, technological change can alter the preconditions and the opportunities for reinforcement learning since the learning process is contingent on the sufficient availability of the reinforcers. Without sufficient availability, a regular pairing of the stimuli and, therefore, the learning process cannot take place.
3.2. (Social-) Cognitive Learning A theory of human behaviour that is restricted to associative learning, however, may be incomplete. Such an approach would entirely ignore the forms of intentional behaviour that are typical for economic processes (Witt, 1987, p. 121). In fact, there are strong indications for the existence of cause–effect learning mechanisms that work even in the absence of, say, the temporal contiguity of the unconditioned reinforcer and the neutral stimulus. Menger’s definition contains a broader range of goods than associative learning theory would predict. Considering that, it is interesting that Menger specifies the necessary condition – his second – that there must be an objective causal connection between needs and things with respect to temporal contiguity. In Menger’s theory, the relationship between consumption and need satisfaction can be more or less immediate. The so-called ‘first-order’ goods that are used for direct need satisfaction, such as bread for example, are distinguished from ‘second’ and ‘third-order’ goods which are connected in a less obvious way to the satisfaction of needs. A typical example of a nonimmediate cause–effect sequence occurs with goods that are inputs in a production process. This idea can be put into more familiar terms: humans have the unique capacity to invest, i.e. to anticipate future rewards. It is probably this specific ability that enables humans to produce tools for less immediate need satisfaction as well as for further tool production in order to increase the availability of first-order goods.31 In ethnology, it is well-known that humans are the only tool-producing species: Since the paleolithicum, humans apparently not only used tools but became able to produce them. It is true that certain monkey species use objects as if they were tools, and cases have been reported in which they have even refined these objects. Humans, in contrast, additionally produce tools in order to produce tools.32 Whereas monkeys use objects only in specific situations and at specific moments, human behaviour regarding the use of tools is more complex since these tools
An Evolutionary Approach to Consumer Behaviour
19
are kept for further use in the future. Moreover, human tools are not just a prolongation of the body and its organs like the simple objects that monkeys use (Hirschberg, 1996, pp. 84–85). It is this tool-producing capacity that has enabled humans to escape selection pressure.33 Humans apply their cognitive capabilities not only to tool production. As consumers, they can anticipate positive and negative effects, too.34 For instance, in the case of some pharmaceuticals, e.g. antibiotics, a considerable time-lag between consumption and its positive impact exists. When the condition of temporal contiguity is violated, however, there must be a mechanism linking causes to effects in the human mind, independent of temporal contiguity. In contrast to reinforcement learning, cognitive learning provides ‘‘knowledge about the world that is not tied to particular behaviour’’ (Toates, 1998, p. 59). Due to their cognitive capacity, humans are able to expand the number of first-order consumer goods beyond the range that is possible by simple reinforcement learning.35 Subsequently, a closer look is taken at cognitive learning mechanisms in order to address changing consumer knowledge. In examining the question of where this knowledge about the world comes from, special attention is paid to social-cognitive learning. Once more applying an evolutionary perspective to psychology, social-cognitive learning concerning cause–effect relations can be assumed to contribute to increasing overall fitness since it enables humans to take advantage of conspecies’ experiences when reacting to unknown situations. How social-cognitive learning of cause–effect relations actually works and under what conditions it takes place is to be highlighted below. Starting from the truism that only beliefs that are perceived can influence actual consumption behaviour, my investigation proceeds in two steps. In the first step perception is analysed and the main focus is on what gets attention and why. In a second step the conditions are examined under which beliefs are transformed into actual behaviour, i.e. become ‘tight’, and how these conditions can change.
3.2.1. The Dynamics of Knowledge (Element 3) The necessity of examining the question of which beliefs receive attention and which do not, arises from the fact that the human capacity to perceive and to process information is bounded. This implies that consumer attention is a scarce resource. Witt (1987, p. 116) notes that only portion of the total amount of information humans are exposed to can be perceived and processed.
20
WILHELM RUPRECHT
Using the metaphor of a camera lens, Witt (1989a) distinguishes two ways in which attention is focussed: The first is shifting the whole focus of attention. Witt calls it the ‘agendasetting effect’. Because retrieval works sequentially, agenda-setting, e.g. by advertisements, is more effective the more often a message is repeated (see also Woo, 1992, p. 97). In modern mass media societies, telecommunication technologies enable the broadcasting of information and centralized agenda setting (Witt, 1996b). The second way to direct attention, which Witt calls the ‘refinement effect’, is to narrow down the focus to one issue which, in turn, can be perceived in more detail. Decentralized communication in small groups, e.g. among hobbyists who continually discuss very focussed issues, can contribute to the refinement effect. Centralized communication and the agenda-setting effect seem to be more relevant than decentralized communication in small groups for an explanation of how beliefs spread from the micro to the macro level – given the corresponding technological preconditions for spreading information centrally. In a mass media society that is characterized by many different radio and TV channels and information overload, however, it is not so easy to assess who actually sets the agenda. Witt (1989a) proposes applying the concept of competition to this issue: competition between different providers of information for the scarce resource ‘consumer attention’ is ongoing. By discussing ‘prominence’ Witt (1989a) elaborates these concepts. While it is often assumed that prominence results from certain unique attributes or combinations of attributes that give a certain item or person a position of natural monopoly, the present approach assumes that prominence can be the result of incidence and historical contingency. Whether beliefs or persons become prominent, and which topics receive attention depends on the relative strength and frequency of information. The recent success of ‘Big Brother’ TV shows in Germany and in other countries and the subsequent commercial exploitation of the prominence gained by several participants in the programmes may support the view that prominence is probably not caused by unique personal characteristics but rather by the fact that the audience has been repeatedly exposed to certain persons or issues. 3.2.2. The Dynamics of Tightness (Element 2) In general, people do not believe every message they receive. How do consumers manage to decide which beliefs to trust and which not, given that objectivity is not a valid criterion? To answer this question, Mokyr’s concept
An Evolutionary Approach to Consumer Behaviour
21
of ‘tightness’ is useful. It can be related to psychological theory. Promising candidates for effecting such linkages are ‘consistency theories’ which start with the premise that people do not necessarily act upon their beliefs. Moreover, consistency theories provide an answer to the question: under what conditions are people inclined to act upon beliefs. A common feature of consistency theories is their postulate that individuals have an innate human need for consistency between the elements of a cognitive system. Since discrepancies between these elements cause a feeling of discomfort, humans try to reduce this discomfort. There are several ways of coping with cognitive dissonance (Zimbardo, 1992, p. 580). For instance, the relevance of new information for an individual’s own behaviour can be systematically ignored, or the information source can be brought into discredit.36 The consistency concept allows questions to be addressed that are usually neglected by the literature on diffusion, in which authors have often relied on personality factors to explain individual differences in the inclination to be innovative.37 Consistency theory, however, involves cultural influences and cultural diffusion constraints. In order to become relevant for behaviour, a new behavioural option must not only be perceived, it must additionally fit pre-existing norms and mental structures. It is true that this has a conservative implication. However, consistency theory does not imply that innovative actions are not considered; it only specifies the conditions under which they are likely to occur. Like Mokyr’s tightness concept, consistency theory presupposes that the diffusion of beliefs or better, the transformation of knowledge into actual consumption behaviour, is contingent on social norms.38 Seen from the point of view of evolutionary economics, consistency theory, like reinforcement theory, has a static component. This can be maintained since it is not capable of explaining the transition from cognitive dissonance to cognitive consonance although it contains different strategies to solve the feeling of discomfort that is caused by dissonance. To put it in Mokyr’s (1999a) terminology, how beliefs which used to be not ‘tight’ become ‘tight’ and vice versa remains an open question. To answer this question, a dynamic theory of tightness is required. Since tightness is a matter of social norms, one would have to recombine consistency theory with a theory of changing norms39 in order to get a dynamic theory of tightness. As long as a group’s opinion leaders do not change their attitudes there will always exist a potential source of cognitive dissonance for the other group members. When they change their attitude, or when other opinion leaders take their place, formerly dissonant knowledge has a chance to become consonant. The dynamics of norms, then, has at least the same relevance for explaining
22
WILHELM RUPRECHT
changing consumer behaviour as is the case for the dynamics of knowledge or beliefs. As the case of Galileo illustrates, such a dynamics of formerly not tight beliefs becoming tight, may take considerable time.
4. CONCLUSIONS My aim in this paper was to provide an evolutionary framework for the analysis of changing consumption patterns that highlights the role of evolving consumer wants as a driving force. As a starting point Carl Menger’s theory of goods was chosen. Focusing on the changeability of user value, a connection between Menger’s 19th-century theory of goods and 20thcentury learning theories was established. A look at Menger’s four conditions for how a thing is to become a good from the perspective of these psychological theories revealed that only consumer needs and consumer knowledge are necessary conditions while a thing’s availability and an objective cause–effect relationship between a thing and need satisfaction are not. The question of how a thing is to become a good, however, should not be confused with the question of how consumption patterns change. While the former refers to the level of an individual consumer, the latter, by contrast, refers to phenomena that occur at a more aggregate level. In order to bridge the gap between the individual and the aggregate level, a perspective considering cultural evolution as being based on biological evolution was applied. Here, however, I suggested that not the content but rather the structure of cognitive and non-cognitive learning processes is genetically coded and common to all humans. In psychology, moreover, it is wellknown that certain conditions – already hinted at in Menger’s theory of goods – are necessary for learning to take place. Both aspects can be exploited to solve the aggregation problem. If (a) the structure of learning processes is genetically coded and (b) learning processes are contingent on certain environmental conditions, then identical environmental conditions should lead to identical learning processes. In the case of non-cognitive reinforcement learning, the sufficient physical availability of reinforcers is the relevant environmental condition. And whether social cognitive learning takes place, i.e. whether consumer knowledge spreads within a population, crucially depends on the prevailing social norms. The physical availability of reinforcers itself, however, is subject to change as well as the prevailing norms. While this physical availability alters as a concomitant of technological change, norms alter in the course of institutional change.
An Evolutionary Approach to Consumer Behaviour
23
By taking into account that the physical and social conditions on which learning takes place are dynamic, my approach not only stresses the adaptive and contingent character of both learning mechanisms, it also allows for an analysis of collective learning processes as historical events and makes reinforcement learning theory and consistency theory fruitful for the analysis of economic change.40
NOTES 1. This characteristic makes Menger’s definition different from either Lancaster’s (1966a, 1966b, 1971) or Saviotti’s (1996) definitions of a consumer good (for the former see Ruprecht (2005), for the latter Appendix A). 2. Unlike Von Hayek (1979, pp. 44–45) who edited Menger’s collected work, however, Menger did not explicitly use the term ‘concept’ in the context of consumer goods. 3. See Herrmann-Pillath (1996). 4. This requires an analytic framework that does not assume an objective ‘production relationship’ between things and ‘need satisfaction’ as proposed, for example, by Saviotti (1996). In his approach, the so-called technological characteristics have no capability to change independently of the so-called service characteristics (see Appendix A). 5. A similar case is described in Mokyr (1998a) who uses an example from the beginning of the 19th century when Louis Pasteur detected the causal connection between household cleanliness and child mortality. Although this connection had always existed, the new knowledge contributed to remarkable changes in hygienic and consumption behaviour. 6. It is useful in this context to separate the generation of new beliefs about cause– effect relationships, be they true or false, from their diffusion. For this purpose, furthermore, objective and subjective novelty can be distinguished. ‘‘There are genuine novelties which have not been previously experienced by anybody including the scientific observer. On the other hand, something may be a novelty in the sense defined above for some particular individual whereas it is already well-known to others, e.g. researcher observing the diffusion of an innovation in a certain region or population’’ (Witt, 1989b, p. 420). 7. The case is taken from Rogers (1995, pp. 1–5). 8. The possibility that placebos may have health impacts which are caused by an individual’s subjective conviction of a positive cause–effect relationship is not denied. 9. There are good reasons, for instance, to be convinced that things containing calories are food items. 10. While Mokyr (1998a, p. 19) has characterized this problem as a ‘problem of induction’, I will – in less philosophical terms – refer to it as a ‘knowledge problem’. 11. It is, of course, by no means excluded that scientifically tested knowledge is simultaneously tight knowledge. 12. Such a view is represented by Kenneth Galbraith (1958, p. 156), who states that ‘‘wants are dependent on production. It [the view in question] accords to the
24
WILHELM RUPRECHT
producer’s function both of making the goods and of making the desires for them. It recognizes that production, not only passively through emulation, but actively through advertising and related activities, creates the wants it seeks to satisfy’’. 13. Rather than being a necessary condition for a thing to become a good, the ‘tightness’ of consumer beliefs seems rather to be a necessary condition for its diffusion. 14. Scarcity can be defined as the case in which the quotient of requirements to availability is larger than 1 (Menger, 1950). 15. This can be stated even though assessing when an individual considers a consumption item to be available given his or her monetary income is an unsolved problem in economic theory. Only little effort has been made so far to find out what determines the income elasticity and the price elasticity of goods. However, it is quite plausible to assume that consumer knowledge and need intensity, i.e. Menger’s conditions 1 and 3, are relevant for such an explanation. 16. In the case of scarce goods, robbery and theft are also possibilities to gain command of a thing, without income. 17. Swann (2001) has argued that in fact Rolls Royce cars are positional goods since they are produced in limited editions. 18. This point I owe to Reinoud Joosten. 19. It is worthwhile to note that this qualified definition only refers to the user value of goods; their exchange value is not taken into consideration at all. 20. It is interesting to note that the analysis of demand-side processes has been completely neglected in the technology push, versus demand-pull debate (for a survey see Mowery & Rosenberg, 1979). 21. Hayek (1979, p. 48). 22. For a similar view see also Vanberg (2004). 23. See, for example, Barkow et al. (1992) or Baker et al. (1996). The idea that the design of behavioural dispositions and the design of learning processes is the result of different adaptations to the physical world is called, in the terminology of recent evolutionary psychology, the ‘modularity’ of human behaviour. 24. In doing so, I do not claim completeness, since it is not clear how many different learning mechanisms may have evolved during human phylogeny. 25. See Witt (2001). While it is assumed that the biological evolution of the cognitive apparatus of homo sapiens was completed during the pleistocene era around 100,000 years ago (see Miller & Todd, 1994), other, more primitive, behavioural dispositions may have evolved earlier, in prehistoric times. 26. Operant conditioning means that behaviours, which are already present in the behavioural repertoire, are repeated more often when they are rewarded (Witt, 2001). 27. In many consumption-related fields, children have no capability of knowing what is good. Hence, parents vicariously expose their children to goods they like themselves. Through reinforcement processes, wants that are acquired by one generation can be transferred to the next. Ceteris paribus, this kind of ‘exposure learning’ implies a genealogical process, which can be compared with the corresponding processes in biology. 28. I am aware that the term ‘preferences’ is not entirely suitable in the present context since it is defined as a relational concept.
An Evolutionary Approach to Consumer Behaviour
25
29. Already the economist Georgescu-Roegen (1954) proposed a ‘theory of wants’ which could be interpreted as a behaviouristic interpretation of his own principle of the ‘irreducibility of wants’, i.e. the variety of wants cannot be reduced to an aggregate of utility, and of the principle of the ‘growth of wants’, i.e. there is always a next want. 30. The Pavlovian dog, for instance, established a relation between the sound of a bell and food. 31. In principle, the same necessary conditions must hold for higher-order goods as for first-order goods. That objectivity in the cause–effect relationship is not a necessary condition becomes plausible when one considers that objectivity refers to the relation between the higher-order good and need satisfaction and not to the relationship between a fourth-order and a third-order good. 32. In anthropology, this observation has inspired a discussion on whether tool production and usage has co-evolved with language (Gibson & Ingold, 1993). 33. This has happened in spite of the scepticism of Thomas Malthus, who denied the possibility of intense growth since he was convinced of the dominance of biologically determined (reproductive) urges over human cognitive and innovative capacities. Through the prolongation of the production chain, the availability of first-order goods can be enhanced. Beside process innovations, simple investment in higher-order goods that are required to produce first-order goods, can also increase the physical availability of first-order goods. 34. The often-made distinction between consumption and production goods according to their positions in the value-added chain is somehow arbitrary. It can for instance be argued that cars may be used as consumption goods as well as in a production process. Stigler and Becker (1977) have argued in their household production theory, quite similarly, that consumers can also produce. With regard to different types of learning processes, therefore, it seems to be more convenient to talk about more or less immediate cause–effect relationships than about consumer and production goods. 35. In the context of cognitive learning and the ability to anticipate cause–effect relations, it is interesting to note that Menger’s academic teacher, Wilhelm Roscher, distinguished humans from other animals according to the number and quality of their wants. According to Roscher (1886, p. 1), the desire for clothes or fire are examples for wants that are specific to the human species. 36. Lauer (1996, pp. 155–165) asserts that consistency theories can be divided into theories that hold for the ‘post-decision phase’ and others that hold for the ‘predecision phase’. An exponent of the latter category is Ajzen’s (1988, 1991) ‘theory of planned behaviour’ which has some relevance for the adoption of innovations when these are understood as actions that have not been carried out earlier (for this definition see Witt, 1993, p. 92). The message of Ajzen’s theory can be summarized as follows: people execute a certain behaviour when, firstly, they themselves evaluate it positively, secondly, when they have the confidence to perform it, and, thirdly, when they believe important people think they should execute it. For the sake of simplicity, subsequently, I refer to the theory of planned behaviour as ‘consistency theory’. 37. For example, Rogers (1995) distinguishes different types of adopters, e.g. ‘early adopters’, ‘later adopters’ and ‘laggards’. Likewise, Witt (1996c) stresses the relevance of differences in the degree of personal curiosity for the inclination to adopt new items.
26
WILHELM RUPRECHT
38. In his explanation of why the industrial revolution started in Britain, Mokyr (1999b, p. 4) directs attention to the possibility that diffusion constraints have a cultural base. Likewise, Rogers (1995) assigns a central role to social authority and opinion leadership. In the case introduced earlier of the ‘boiling water practice’ in the Peruvian village, he conjectures that the diffusion agent suffered from a lack of social reputation. 39. See e.g. Witt (1989c). 40. Along these theoretical lines, a case study on the development of the consumption of sweeteners has been conducted. Changes in food consumption patterns over time were explained as the outcome of collective learning processes. As they have been contingent on systematically changing environmental conditions, these learning processes can be considered as historical events. It turned out, moreover, that the presented approach complements the Lancasterian characteristics approach to the adoption of novelty in consumption (see Ruprecht, 2005). 41. For a comprehensive survey of this approach, which aims at explaining how artefacts emerge and how they evolve, see Saviotti (1996). 42. In this respect it resembles Kelvin Lancaster’s (1966a, 1966b, 1971) ‘indirect utility approach’ which also is based upon a production metaphor (for a discussion see Ruprecht, 2005). 43. Elsewhere, Saviotti mentions a ‘pattern of imagination’, represented by the double arrow between technological and service characteristics, which however, is not explained further.
REFERENCES Ajzen, I. (1988). Attitudes, personality, and behavior. Chicago: Dorsey Press. Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50, 179–211. Baker, A., Murphy, R., & Valee-Tourangeau, F. (1996). Associative and normative models of causal induction: Reacting to versus understanding cause. The Psychology of Learning and Motivation, 34, 1–45. Barkow, J., Cosmides, L., & Tooby, J. (Eds). (1992). The adapted mind. Evolutionary psychology and the generation of culture. New York: Oxford University Press. Bourdieu, P. (1984). Distinction. A social critique of the judgement of taste. London: Routledge & Kegan Paul. Galbraith, J. K. (1958). The affluent society. Boston: The Riverside Press. Georgescu-Roegen, N. (1954). Choice, expectations and measurability. Quarterly Journal of Economics, 68, 503–534. Gibson, K., & Ingold, T. (Eds). (1993). Tools, language and cognition in human evolution. Cambridge: Cambridge University Press. Gru¨nbaum, A. (1984). Explication and implications of the placebo concept. In: G. Andersson (Ed.), Rationality in science and politics (pp. 131–158). Dordrecht: Reidel. Hayek, F. A. (1979). The subjective character of the data of the social sciences. In: F. A. Hayek (Ed.), The counter-revolution of science (pp. 41–60). Indianapolis: Liberty Press. Herrmann-Pillath, C. (1996). On the ontological foundations of evolutionary economics. Mimeo. Hildenbrand, W. (1994). Market demand. Princeton: Princeton University Press.
An Evolutionary Approach to Consumer Behaviour
27
Hirsch, F. (1976). Social limits to growth. Cambridge, MA: Harvard University Press. Hirschberg, W. (1996). Hand und Gera¨t – Der Weg zu einem Symbol. In: M. Liedtke (Ed.), Kulturethologische Aspekte der Technikentwicklung (pp. 82–90). Graz: Austria Medien Service. Lancaster, K. (1966a). Change and innovation in the technology of consumption. American Economic Review, 56, 14–23. Lancaster, K. (1966b). A new approach to consumer theory. Journal of Political Economy, 74, 132–157. Lancaster, K. (1971). Consumer demand. New York: Columbia University Press. Lauer, T. (1996). Die Dynamik von Konsumgu¨terma¨rkten. Heidelberg: Physika. Menger, C. (1950). Principles of economics. Glencoe, IL: The Free Press. Millenson, J. (1979). Principles of behavioural analysis. New York: MacMillan. Miller, G., & Todd, P. (1994). A bottom-up approach with a clear view of the top: How human evolutionary psychology can inform adaptive behaviour research. Adaptive Behaviour, 3, 83–95. Mokyr, J. (Ed.). (1998a). Technological selection, information, and changing household behavior, 1850–1914. In: Neither chance nor necessity: Evolutionary models and economic history. Princeton: Princeton University Press. Mokyr, J. (1998b). Induced technical innovation and medical history: An evolutionary approach. Journal of Evolutionary Economics, 8, 119–137. Mokyr, J. (1999a). Demand as a factor in the industrial revolution. Northwestern University, Evanston, mimeo. Mokyr, J. (1999b). Knowledge, technology, and economic growth during the industrial revolution, Northwestern University, Evanston, mimeo. Mowery, D., & Rosenberg, N. (1979). The influence of market demand upon innovation. Research Policy, 8, 103–153. Nordhaus, W. (1998). Quality change in price indices. Journal of Economic Perspectives, 12, 59–68. Pulliam, H., & Dunford, C. (1980). Programmed to learn: An essay on the evolution of culture. New York: Columbia University Press. Rogers, E. (1995). Diffusion of innovations. New York: The Free Press. Roscher, W. (1886). Grundlagen der Nationalo¨konomie. Ein Hand- und Lesebuch, Stuttgart: Cotta. Ruprecht, W. (2005). The historical development of the consumption of sweeteners – a learning approach. Journal of Evolutionary Economics, 15, 247–272. Saviotti, P. (1996). Technological evolution, variety and the economy. Cheltenham: Edward Elgar. Skinner, B. (1966). Operant behaviour. In: W. Honig (Ed.), Operant behaviour – areas of research and application (pp. 12–32). New York: Meredith Corp. Stigler, G., & Becker, G. (1977). De Gustibus Non est Disputandum. American Economic Review, 67, 76–90. Swann, P. (2001). The demand for distinction and the evolution of prestige car. Journal of Evolutionary Economics, 11, 59–75. Toates, F. (1998). The interaction of cognitive and stimulus–response processes in the control of behavior. Neuroscience and Biobehavioural Reviews, 22, 59–83. Vanberg, V. (2004). Austrian economics, evolutionary psychology and methodological dualism: Subjectivism reconsidered. In: R. Koppl (Ed.), Evolutionary Psychology and Economic Theory, Advances in Austrian Economics, 7, 155–199.
28
WILHELM RUPRECHT
Witt, U. (1987). Individualistische Grundlagen der evolutorischen O¨konomik. Tu¨bingen: Mohr. Witt, U. (1989a). Wissen, Pra¨ferenzen und Kommunikation – eine o¨konomische Theorie. Analyse und Kritik, 11, 94–109. Witt, U. (1989b). Subjectivism in economics: A suggested reorientation. In: K. Grunert & F. O¨lander (Eds), Understanding economic behavior (pp. 409–431). Dordrecht: Kluwer. Witt, U. (1989c). The evolution of economic institutions as a propagation process. Public Choice, 62, 155–172. Witt, U. (1993). Emergence and dissemination of innovations: Some principles of evolutionary economics. In: R. Day & P. Chen (Eds), Nonlinear dynamics and evolutionary economics (pp. 91–100). Oxford: Oxford University Press. Witt, U. (1996a). Innovations, externalities and the problem of economic progress. Public Choice, 89, 113–130. Witt, U. (1996b). The political economy of mass media societies. Papers on economics and evolution no.9601, Max Planck Institute, Jena. Witt, U. (1996c). A ‘Darwinian revolution’ in economics? Journal of Institutional and Theoretical Economics, 152, 707–715. Witt, U. (2001). Learning to consume. Journal of Evolutionary Economics, 11, 23–36. Witt, U. (2003). Evolutionary economics and the extension of evolution to the economy. In: U. Witt (Ed.), The evolving economy (pp. 3–34). Cheltenham: Elgar. Woo, H. (1992). Cognition, value and price. Ann Arbor: The University of Michigan Press. Zimbardo, P. (1992). Psychologie. 5. Auflage. Berlin: Springer.
APPENDIX: THE SAVIOTTIAN CONCEPT OF AN ARTEFACT According to Saviotti’s and Metcalfe’s twin approach,41 artefacts consist of two sets of characteristics – technical and service – which very much resembles the distinction between genotype and phenotype in evolutionary biology. In Figure A.1, the two tupels represent n means (X) and m ends (Y), i.e. the entities which make up the efficiency relation of an artefact i.
Fig. A.1.
Technical Characteristics
Service Characteristics
Xi1
Yi1
Xi2
Yi2
Xin
Yim
The Twin Characteristics Representation of Artefact i. Source: Saviotti (1996, p. 64).
29
An Evolutionary Approach to Consumer Behaviour
Technical characteristics are connected to service characteristics by a kind of production relationship.42 Saviotti (1996, p. 64) compares technical characteristics with buttons ‘‘y that, when pushed, produce required services’’.43 For illustration purposes, Saviotti (1996, p. 67) chooses the example of helicopters. The ‘length of a diameter’ is a technical characteristic that produces the service characteristic ‘maximum take off power’. Other examples are the technical characteristic ‘engine power’ that produces the service characteristic ‘maximum speed’, or the technical characteristic ‘number and geometry of engines’ that produces the helicopter’s range as a service characteristic. There is of course, an essential difference between the twin approach and the Mengerian concept of a good. Since the two types of characteristics are connected by a production relationship, service characteristics can change only when technical characteristics have changed first (see Table A.1). Since the Saviottian approach relies on a unimodal ontology, in contrast to Menger’s approach, an independent change of service characteristics is not possible. Table A.1.
Changing Characteristics without Subjectivity in Saviotti’s Approach. Identical service characteristics
Identical technological characteristics Technological characteristics change
Service characteristics change
Identical good
Not defined
Cost reducing ‘process’ innovation, intertechnological competition
Entirely new product
This page intentionally left blank
WHAT SHALL I DO? (OR WHY CONSUMER THEORY SHOULD FOCUS ON TIME-USE AND ACTIVITIES, RATHER THAN ON COMMODITIES)$ Ian Steedman ‘y do not squander time, for that’s the stuff life is made of.’ Benjamin Franklin, The Way to Wealth, 1757
INTRODUCTION The sole purpose of this simple paper is to present the idea that theorists of consumption, orthodox or otherwise, might do well to focus their attention on the use of time. ‘Economics is at bottom the study of how humans spend their lifetimes’ (Georgescu-Roegen, 1983, p. lxxxv), after all, and it thus makes sense to place time-use at centre-stage and to make sure that it is considered explicitly within consumer theory. Such an emphasis, it will be $
An earlier version of this paper was presented to the Gaeta workshop (18–20 March 2005) on Economic Theory and the Practice of Consumption. Evolutionary and Other New Approaches.
The Evolution of Consumption: Theories and Practices Advances in Austrian Economics, Volume 10, 31–40 Copyright r 2007 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 1529-2134/doi:10.1016/S1529-2134(07)10002-8
31
32
IAN STEEDMAN
urged, enables the economic theorist to connect more easily both with certain other social-scientific and philosophical concerns and with many everyday common-sense concerns. ‘What shall I do?’, for example, is both a more frequent and a deeper question than, ‘What shall I buy/consume?’ (‘What ought I to be?’ is no doubt a still deeper – but less frequent! – question but is too difficult to be considered here.) To place the use of time at centre-stage is at once to introduce into consumer theory an element common to all consumers, to supplement (counteract?) the standard individual, idiosyncratic emphasis. No matter what incomes and preferences may be, everyone has exactly 10,080 min per week to ‘spend’ in one way or another. The ‘time constraint’ is inevitably the same for everyone, however different their budget-constraints may be. (It should perhaps be emphasized that the time-constraint on time-uses is, unlike the budget constraint, an identity and not a mere weak inequality.) To say this is not, of course, to imply that incomes, prices and commodity quantities become unimportant; they do not. Nor is it to suggest that no traditional concepts – e.g., Hicksian complements, or inferior goods – will continue to be important; some do. (Indeed, the time-constraint identity immediately gives great force to the concept of opportunity cost.) As Schopenhauer once remarked, ‘Buying books would be a good thing if one could also buy the time in which to read them’. This witty (or resigned?) aphorism may serve to introduce the fact that, at least for ‘moderns’ who buy their books already cut, the use of time to which we refer is ‘pure’ consumption time. We shall not be concerned with Becker-like ‘preparationfor-consumption’ time (significant as that may be in various contexts). We shall suppose throughout that the intelligent theorist is concerned to understand human action, even when understanding does not permit prediction. (Just as geophysicists understand earthquakes, even though their understanding yields little or no ability to predict the timing, frequency or magnitude of these phenomena.)
SATIATION If your income were sufficiently large (relative to all the relevant prices), you could spend the next period of time as you wished (unconstrained by income, that is – but not of course by ethical commitments, social obligations, etc.). You could, that is, adopt your ‘satiation’ time-use pattern, which would entail in turn the use of some particular set of commodities. In the absence of the budget constraint there would always be a ‘satiation
Why Consumer Theory Should Focus on Time-use and Activities?
33
commodity bundle.’ (The only exceptions would relate to commodities whose ‘use’ required no time; they are ignored here, for simplicity. The conventional preference map over commodity bundles implicitly assumes, by contrast, that commodities can be consumed at an infinite time-rate.) With a fixed set of uses of time, fixed commodity inputs per unit of activity and fixed prices, the growth of income would eventually bring one to the position described above; commodity satiation would be inevitable (eventually) and everyone would have income they could not spend. Why do we not often observe this? Producers generate a constant stream of new uses of time and new commodities to go with them – but have not yet found a way to give anyone more than 168 h a week, so that new time-uses often replace old ones. It would thus be of interest to develop a ‘growth without satiation’ theory that centres on the time constraint and on our changing ways of using time. Note that producers often have an incentive to promote commodityintensive uses of time, whether or not some commodity – unintensive uses would be good for us.
THE TIME-CONSTRAINT AND CONVENTIONAL RESULTS It has been shown that a number of familiar results in conventional consumer theory are changed by the presence of the time-constraint (Steedman, 2001, especially Chapters 1 and 3). Before mentioning some of them, we should perhaps remark that this does matter to consumption theorists who have turned their backs on the conventional theory. This is so, first, because some broadly conventional concepts and results are likely to re-emerge in the context of non-orthodox theory. And it matters, secondly, because mainstream theorists will be the more open to unconventional approaches, the more flaws are found to exist within the standard approach. In both these respects, then, criticism of orthodox theory is a contribution to innovatory theory, albeit an indirect contribution and one not producing instant results. Pazienza! To keep matters simple, consider a three-commodity analysis in which each commodity takes a fixed amount of time (per physical unit) to be consumed; suppose that no other use of time exists. Then, in obvious notation (t1 x1+t2 x2+t3 x3 T). Note that this is an identity; it would have no meaning to speak of using 167 or 169 h in a week (say). So, not only does the consumer face two constraints (time and money income) but it is only over (x) bundles lying on the time-constraint that it is meaningful for the
34
IAN STEEDMAN
consumer to have a preference ordering. The analytical framework is at once markedly different from that of the textbook. If the budget-constraint is such that the consumer cannot achieve the satiation bundle, how will the bundle purchased respond to changes in income and prices? Inferior goods easily arise, Giffen goods are readily found and, more generally, ‘any reasonably simple or useful conclusions about [@xi/@pj]’ are hard to come by; Hicksian complements become more likely, as does an inverse relation between hours worked and the real wage rate (Steedman, 2001, pp. 66–68). If we hope that more definite results might emerge when looking at uses of time as a function of their costs (i.e., the aggregate costs of the commodities they involve), then we shall be disappointed. There can be ‘Giffen time uses’, there must be ‘Hicksian time-use complements’ and there must be at least one ‘income inferior time-use’ (2001, Chapter 3). Non-convex consumption sets can arise, as can discontinuities in commodity demand curves. The simple fact that consumption takes time thus makes a considerable difference to conventional results – and it is a disturbing difference, giving yet more reason to be dis-satisfied with orthodoxy and thus to be ready to consider alternatives.
GENERALITY A simple version of time-centred consumer theory includes the standard theory as a special case – and does so without involving any great analytical complexity. Consider first, a one-period analysis and let the column vectors (z, y, t, x) represent characteristics a` la Lancaster, an abstract ‘activity’ vector, time-use and commodity quantities, respectively; let the row vectors (p, s) be vectors of commodity prices and of unit elements, respectively; let the scalars (e, T) represent expenditure and total time in the period. We formalize the consumer’s decision problem as Max
subject to
uðzÞ 8 > > > > > > < > > > > > > :
z ¼ Ay t ¼ By x ¼ Cy
px e st T and y 0
where the meaning of matrices A, B and C will be clear. (Note, though, that proportional columns in C can be used to represent alternative rates of
Why Consumer Theory Should Focus on Time-use and Activities?
35
consumption; y17 and y18, for example could be ‘slow beer and pizza consumption’ and ‘fast beer and pizza consumption’ activities, with C17 and C18 proportional to one another. Whether ‘faster’ is better or worse than ‘slower’ can depend both on preferences and on the circumstances.) Setting C I and dropping both the st T identity and the equation t ¼ By would yield a textbook model, of course, so that the above model is more general – both in recognizing the time constraint and in allowing for different rates of consumption. Secondly, just as conventional consumption theory can be extended to yield a multi-period analysis, so the above (z, y, t, x) model can be elaborated to represent a consumer’s choices over successive time periods. (This is naturally important for issues of, for example, habit formation, learning, novelty, addiction a` la Elster, etc.) Of course, the earlier mentioned results concerning inferior and Giffen goods (or time -uses) continue to hold, as does the greater likelihood (in some cases, certainty) of Hicksian complementarity. The reader will see at once, without need of explicit elaboration, how such results can affect one’s analysis of the effects on consumer demand of increases in income tax, reductions in ad valorem or excise taxes, of changes in tariffs, etc. If definite predictions become harder to obtain, that may simply reflect the way the world is. Our job is to understand, not only to predict. The obvious way to analyse this model is clearly to eliminate z, t and x at first; but this naturally does not mean that we are prevented from studying (@xi/@pj), (@xj/@e), etc. The same is true when we move on to a multi-period model (Steedman, 2001, pp. 107–113), or when we recognize that T may be endogenous, depending on diet, exercise, etc. (Steedman, 2001, pp. 132– 134). By comparison with the standard commodity-centred analysis, then, the time-centred one is both more general and permits the ready recognition of important real-world considerations (such as rates of consumption and the presence of the time-constraint). Our remarks thus far have been of a rather general nature, so we now turn to offer a few more specific illustrations of areas in which the time-centred focus would seem to be advantageous. In each case the argument will only be briefly sketched but may suffice both to make the point and to provoke the reader to further reflection.
GIVING AND SHARING OF TIME AND OTHER INTRINSICALLY VALUED ACTIVITIES It goes without saying that time given to childcare, or to care of the elderly, or to voluntary and charitable activities of various kinds, is a significant part
36
IAN STEEDMAN
of many people’s lives. So are friendship and various other time-sharing activities. It is well known that the ‘giving’ of one’s time is sometimes far more appreciated than the giving of money. It will be clear whether these things fit more readily into a time-based or a commodity-based theory of household behaviour. And it ought to be clear that this matters, even when one’s main goal is to understand the consumption of commodities, since that consumption may be heavily influenced by a household’s ‘wider’ concerns. Activities, uses of time, in general (and not just the sharing of time) are perhaps distinguished more readily than are commodities into those valued for their own sake, those valued only instrumentally, those valued in both ways, and those valued both positively and negatively. And ethical questions certainly fit far more naturally into a ‘What shall I do?’ framework of thought than into one centred on ‘What shall I consume?’ Since any reasonably normal consumer does in fact value some activities for their own sake, does have mixed feelings about some activities and does have ethical concerns; these facts provide further reasons for setting one’s theory of consumption within a time-use framework – this, of course, being perfectly consistent with the intention to study the use of commodities. The same is true of that particular aspect of ethical life that involves acting ‘now’ in such a way as to change one’s own dispositions in the future; while such action may involve the use – or avoidance – of certain commodities, the emphasis more naturally falls on what one is doing (or refraining from doing). Consider, for example, the case of someone seeking to become more generous. The process will have much more to do with the practice of (and perhaps the avoidance of) certain ways of acting, than it will with specific commodities. Commodities may, of course, be involved (but need not necessarily be, since the giving of time is important here) and, when they are, it may matter in any particular act of generosity just which commodities are given. It will, nevertheless, be the kinds of actions engaged in (or refrained from) that are central to developing generous dispositions. In such a case, moreover, the distinction between valuing such actions for their own sakes and valuing them for their tendency to strengthen the generous disposition may become somewhat blurred.
BRIDGE BUILDING A time focus will not only make it easier for the economist to converse with the moral philosopher. It will also aid discussion with the practitioner of
Why Consumer Theory Should Focus on Time-use and Activities?
37
time-budget studies, with the sociologist, with the anthropologist, etc. Economists using a time-centred approach will find it easier to make themselves understood by others. It is worth noting that the ‘simple’ matter of two people going for a walk together (sharing time) has been presented by a serious political philosopher as a paradigmatic example of social phenomena, since it involves mutual commitment and shared beliefs and goals (Gilbert, 1966). In so far as Gilbert is on to something here, there is again much to be gained if economists focus on the use of time. Even if it be true that, contrary to hostile critics, commodity-centred consumer theory is not intrinsically ‘individualist’, ‘atomistic’, etc., a time-centred approach can perhaps help to make the ‘social’ nature of the economist’s consumption theory more transparent. Intra-economics discussion may also be facilitated by a time focus for consumer theory. When standard welfare theory is presented in terms of Pareto efficiency and preferences over commodity bundles, it is not that easy to conduct a conversation between conventional welfare economics, classical-liberal Austrian theory and Sen’s basic capabilities arguments. Suppose now, however, that the first component is replaced by a timecentred analysis driven by the agent’s question, ‘What shall I do?’ The new ‘trialogue’ starts from a much less awkward basis, since all three elements can now be formulated in terms of what one wishes to do, or should be allowed to do, or is capable of doing. Consider, for example, a debate about whether education and training for adults ought to be financed and/or organized by the public authorities. There will still be plenty of room for disagreements, of course, but at least there will be a (tolerably) common language within which they can be expressed and mutually comprehended.
DECISIONS AND CONVENTIONS We have already suggested above that a ‘What shall I do?’ focus for consumer theory may be more amenable to the explicit recognition of personal obligations and social conventions as factors affecting everyday consumer choices. It will not come amiss, then, to note that at the ‘very beginning’ of time-centred consumer theory, Gossen (1983 [1854]) already stressed the importance of social custom in influencing consumption choices. Social conventions, of course, help to delimit the sets of expectations one needs to consider but Matthews (1991, p. 114) has noted that the very fact of thinking in terms of, ‘What shall I do?’ may itself ‘downgrade the importance of
38
IAN STEEDMAN
expectations.’ When the motivation of action is not purely consequentialist – and in the extreme, if an ‘activity is purely an end in itself’ – then expectations may become less (but not necessarily un-) important. More generally, decision making can take time (even much time) and, as Linder (1970, pp. 61–67) has observed, a time-centred analysis, with its inevitable emphasis on time-costs, provides a natural setting for the recognition of habits, satisficing, bounded and procedural rationality, etc. within consumer theory. It can allow too for the need to ‘keep one’s diary open’, to allow for future flexibility (liquidity) in one’s use of time. There is no good reason to confine consideration of such matters to the theory of the firm! (Quite the contrary, if firms are under more pressure to ‘optimize’ than are households.)
HABIT FORMATION, LEARNING, PRACTICING Consciously acting ‘now’ in such a way as to change my future self, referred to above in an ethical context, need not be an essentially ethical matter, of course; I may be learning, or practicing, or training myself in some way or the other. Such activities may require the purchase and consumption of commodities, but they are still most naturally thought about in terms of ‘What shall I do?’ Even habit-forming activities (and consumptions) that are not engaged in with the express intention of creating/reinforcing a habit can still fit readily into time-centred analysis and the growing literature on addictions of various kinds might well benefit from the use of such analysis. Consideration of such matters will best be undertaken in a multi-period analysis, of course. Present experience may introduce one to ‘new’ uses of time, to the existence of previously unknown characteristics, or to a revised understanding of the numerical values of the elements in Aj. Moreover, Cj may depend on one’s skill in consumption and thus on previous experience; acquisition of skills itself takes time (and may or may not be an intrinsically enjoyable use of time). In some cases, at least, learning to use one’s time well may be harder than learning to manage one’s monetary budget constraint. Finally – and the final indignity for textbook theory! – even for a given set of characteristics the u(z) relation may change with experience. If this all creates the impression that the poor consumer faces a harder task than ‘Max u(x), subject to px p e’, what of it? That is life. Less flippantly, though, we may wonder how well utility maximization can capture decision taking
Why Consumer Theory Should Focus on Time-use and Activities?
39
when so many current actions have habit-forming and other effects (not precisely known) upon future wants and abilities. (The switch from a commodity-focus to a time-use focus does not ipso facto resolve this problem, needless to say.)
REPETITION AND NOVELTY Most people probably want a large measure of stability and repetition in their lives, together perhaps with some (limited) elements of novelty. Certainly both repetition and novelty are central features of consumption activity in the modern world. Repetition was referred to above, of course, in the reference to practicing, etc. but its importance in consumption activities goes well beyond that; one bowl of muesli each morning is not just the same thing as seven bowls of muesli per week. Timing, frequency, duration of each given ‘bout’ of consumption activity, etc. are all of great importance in everyday life and it is clear that a time rather than a commodity focus is congenial to their analysis – and that it is the multi-period version that is needed. (Repetition with respect to daily, weekly and yearly cycles is often bound up with social customs and conventions, of course.) Novelty, in the present context, may be thought of both in terms of new commodities to be used in (broadly) unchanged activities and in terms of completely new activities (even if the distinction between new activities and ‘largely unchanged’ ones is no doubt vague). As noted above, producers may have an incentive to promote new commodity-intensive uses of time by consumers and new uses of time necessarily compete very sharply with old uses, there being only 168 h per week, come what may. (Think, for example, of all the young men who have abandoned their study of Plato (in the public library), for the thrills of skateboarding.) Again, learning to use time in new ways may be difficult. Insofar as consumers are strongly attached to their present allocations of time, this inexorably sharp time-conflict may, of course, make it difficult for producers to get people to take up new activities (as opposed to new commodity inputs to old activities). Here, ‘fashion’ may be on the producers’ side, if a fashionable activity is one to which those pursuing it have no deep attachment. (More serious, presumably, is the boredom of those whose tight budget constraint prevents them from engaging in most of the activities they would like to practise; and of those who search endlessly for satisfying novelty but never find it.)
40
IAN STEEDMAN
EXHORTATION For all the varied reasons given above, the theorist of consumption – whether mainstream or unorthodox – is encouraged to think about the consumer’s decisions and actions primarily in terms of the use of time. Even when the theorist’s central objective is to understand the consumption of commodities, it may well be better not to place them at centre-stage.
ACKNOWLEDGEMENTS I thank Marina Bianchi, a referee and the workshop participants for their encouraging and interesting reactions.
REFERENCES Georgescu-Roegen, N. (1983). Introduction to Gossen (1983). Gilbert, M. (1966). Walking together: A paradigmatic social phenomenon. In: M. Gilbert (Ed.), Living together. Towota, N.J: Rowman & Littlefield. Gossen, H. H. (1983 [1854]). The laws of human relations, etc. Cambridge, MA: MIT Press. Linder, S. B. (1970). The harried leisure class. New York: Columbia University Press. Matthews, R. (1991). Animal spirits. In: J. G. T. Meeks (Ed.), Thoughtful economic man. Cambridge: University Press. Steedman, I. (2001). Consumption takes time. Implications for economic theory. London: Routledge.
IDIOSYNCRATIC LEARNING, CREATIVE CONSUMPTION AND WELL-BEING Marina Di Giacinto and Francesco Ferrante ABSTRACT The consensus view is that economists should observe consumer choices and abstain from investigating the psychological and physiological causes of wants, or the mechanisms governing the formation of preferences. This may be a correct procedure as far as ordinary functional goods are concerned. Problems tend to arise with creative goods (e.g. cultural goods) whose consumption (i) requires skills acquired through education and experience and (ii) generates positive and negative feedbacks and learning-by-consuming processes. This paper presents a simple model of local learning explaining the idiosyncratic accumulation of consumption human capital. Consumption generates local feedback mechanisms whose characteristics depend on the nature of goods and on the type of agent. The model provides some insights on the microeconomics of creative consumption and on the specific role of education.
For I would almost say that this very thing, self-knowledge, is wisdom and I am at one with who put up the inscription of those words at Delphi (Plato, Charmides, 160 D-E). The Evolution of Consumption: Theories and Practices Advances in Austrian Economics, Volume 10, 41–73 Copyright r 2007 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 1529-2134/doi:10.1016/S1529-2134(07)10003-X
41
42
MARINA DI GIACINTO AND FRANCESCO FERRANTE
1. INTRODUCTION The consensus view of economists is that the formation of wants does not fall within the province of economic analysis. Economists should observe consumer choices to arrive at the underlying preference structure, and refrain from investigating the psychological and physiological causes of wants or the mechanisms governing the formation of preferences. Thus, a preference ordering can always be derived from actual market behaviour, to be incorporated in well-behaved utility functions. A world of convex consumption technologies and perfect information can be considered as an acceptable approximation to reality when we refer to ordinary functional goods. Leaving aside social interactions, problems arise in the case of creative goods since (i) their consumption requires skills and information that can only be generated through education, training and exposure to consumption and (ii) the associated hedonic/affective experience stemming from the consumption of these goods activates idiosyncratic learning processes and generates psychological feedback from consumption back to preferences. A useful microeconomic model of creative consumption should take both of these aspects into account. Creative goods are those traditionally associated with artistic, recreational and entertainment activities, though many functional goods also have a creative dimension. Hence, market for creative goods is larger than generally claimed. Most important, it is expanding rapidly, and in the last two decades, economists’ interest in the subject has increased accordingly. Unfortunately, the building up of convincing microeconomic foundations for creative consumption still lags. In this paper, we take these points and explore the main consequences of adopting a restricted notion of learning based on the idea that the feedbacks from consumption exposure back to preferences are local, i.e. idiosyncratic. That is to say, the physiological and psychological mechanisms activated through consumption are assumed to be specific both to the goods consumed and to individuals experiencing it: as such, the result of learning can be very imperfectly codified and de-codified and used as valuable information. Hence, well-being delivered by creative consumption results from the interaction between the subjective characteristics of individuals and the objective creative content of the consumption goods. The approach raises fundamental methodological questions on how one should build microeconomic models of creative consumption and on the role of information in creative consumption demand. Notably, the theoretical framework we develop shows that, under reasonable assumptions and
Idiosyncratic Learning, Creative Consumption and Well-Being
43
circumstances, consumption history matters and that personal information about the feedbacks from consumption to preferences cannot be exploited by others. This model with the insights it makes possible, carries important implications for the recent debate on the ‘‘economics of happiness’’. It suggests that the absence of a clear link between income and happiness (Layard, 2003) might be due to a substantial part of society being locked in consumption styles characterized by habit and lack of subjective creative content. Building on the idea that learning is idiosyncratic, we suggest that the provision of education and of a wide-ranging early life-exposure to creative activities improves the opportunities for creative consumption and can help people to escape from the tyranny of habituation in later stages. Specifically, we argue that an appropriate mix of specific and general human capital reduces the cost of buying more creative preferences and consumption styles.1 Section 2 provides a brief account of the standard models of endogenous preferences based on habit formation. In Section 3, we develop the model of idiosyncratic learning. The static model of local learning to explain the accumulation of cultural capital has important implications for the economics of happiness. Finally, Section 4 draws the main conclusions.
2. CONSUMPTION ANALYSIS WHEN HEDONIC/AFFECTIVE EXPERIENCE MATTERS The formation of habits due to past consumption experience is now commonly cited instance of consumption feedbacks and forms the natural starting point for a discussion of the accumulation of consumption human capital and the evolution of preferences. Within this research area there are two main analytical strands which reach quite different conclusions on the relevance of this form of path-dependence to economic analysis. The first approach originates from well-known contributions on rational habit formation (Stigler & Becker, 1977; Becker & Murphy, 1988), whereby the evolution of preferences is said to be the result of a non-myopic dynamic optimization process (that is, the agents know the marginal rate of substitution of current for future consumption for any available choice). Hence, they can strategically modify consumption over time to exploit the impact of habit on preferences and utility. In such a way, habit formation looses its peculiar nature of a time-consuming process. The second approach takes a less sanguine view of agents’ awareness and ability to learn the psychological mechanisms governing habit formation
44
MARINA DI GIACINTO AND FRANCESCO FERRANTE
and reaches the opposite conclusion, namely, that contingent choices are affected by the dynamics of preferences in a way agents do not anticipate (Pollak, 1970, 1976, 1978). On this view, agents take their consumption history into account but are not able to recognize the impact of present consumption on future tastes (that is, they do not know the marginal rate of substitution of current for future consumption). The present contribution, based on Ferrante (1999), can be placed within the empty intersection between these two approaches on the role of consumption feedbacks, since it posits that agents are locally rational, that is, they form expectations on the evolution of preferences but that the underlying learning process is centred on the actual consumption choices that they have experienced. Hence, whereas Becker and Murphy’s inter-temporal rationality depends on some form of innate knowledge, local rationality posits only knowledge produced by actual exposure to consumption generates idiosyncratic learning about people’s tastes. Moreover, such knowledge cannot be codified and transferred among individuals, being the result of subjective physiological and psychological processes. In other words, de-codification of other people’s experiences to get valuable information is a very imperfect, unreliable mechanism. On more general grounds, the idea of inter-temporal rationality underlying economic analysis and the postulate of consistency of consumers’ choices have been strongly attacked by psychologists also on the basis of the empirical evidence showing biases in the ability of individuals to predict their future tastes (e.g. Thaler, 1980; Loewenstein & Thaler, 1989; Loewenstein, 1988, 1992; Kahneman & Snell, 1992). Within the psychologist-led research area, prospect theory provides helpful insights into the role of consumption experience and habits, seen as frames that affect preferences and choices. The basic intuition is that people evaluate the hedonic consequences of choices as differences from a reference point (Tversky & Kahneman, 1986, 1990, 1991). In this context, the endowment of consumption experience or memory should be seen as a potential source of reference-dependent preferences.2 This recognition of the connection between short-sighted choices, hedonic consumption experience and long-run outcomes should prompt rethinking of the normative implications of consumption theory. In particular, if one does not consider rational habit formation as a plausible account, the alternative – that learning is locally rational – would undermine the basis of standard welfare economics and demand analysis for creative goods. As Pollak (1978) suggests, in a world characterized by myopic habit formation, the competitive equilibrium does not guarantee that everyone gets what he/she wants but rather the converse; that is, that people come to want what
Idiosyncratic Learning, Creative Consumption and Well-Being
45
they get. In this paper, we suggest that these disturbing results have may have to do with the idiosyncratic nature of learning rather than with myopia per se. The interpretative framework presented here does not pretend to give a full account of the factors affecting consumers’ intertemporal choices. Or to describe a full-fledged dynamic model of idiosyncratic learning. Our far more limited aim is to offer a coherent account of the effects of the hypothesis that the accumulation of consumption human capital through experience is an idiosyncratic process. Building on this, we identify the main factors that affect learning and the resulting patterns of accumulation of consumption human capital. To simplify our arguments, we do not consider explicitly3 the role of social interaction as a source of learning.
3. THE ACCUMULATION OF CONSUMPTION HUMAN CAPITAL AND THE EVOLUTION OF PREFERENCES WHEN HEDONIC EXPERIENCE IS LOCALIZED At the heart of consumer theory, and especially of models of rational habit formation, lies the idea that preferences are exogenous, complete and based on reliable information: ‘‘The economist’s traditional picture of the economy resembles nothing so much as a Chinese restaurant with its long menu. Customers choose from what is on the menu and are assumed always to have chosen what most pleases them. That assumption is unrealistic, not only of an economy, but of Chinese restaurant. Most of us are unfamiliar with nine-tenths of the entre´es listed; we seem invariably to order either the wrong dishes or the same old ones. Only on occasions when an expert does the ordering do we realize how badly we do on our own and what good things we miss. The trouble we have with the economy’s menu stems not only from our lack of skills in ordering, but also from our lack of skill in consuming, from the impossibility of making substitutions [y]. The traditional theory of the consumer’s behavior fails to recognize his need for novelty and variety, his need of consumption skills to enjoy certain forms of consumption, and habit as a force which can prevent satisfaction or rational choice’’ (Scitovsky, 1992, pp. 149–150). In this paper we argue that the veil of ignorance about our likes and dislikes is not symmetrically distributed across the consumption possibilities open to us but is either thin or thick, depending on our stock of
46
MARINA DI GIACINTO AND FRANCESCO FERRANTE
consumption experience. Accordingly, we posit that the information generated by experience is local, i.e. is confined to those goods that are object of hedonic/affective exposure and that personal experience cannot be substituted for by information provided by others, i.e. one cannot de-codify the hedonic and affective experiences of other individuals.4 This logic of local learning is very much akin to the same idea, applied long ago to technical change by Atkinson and Stiglitz (1969). Whereas in the latter, what is affected by local learning is the technical coefficients of the production function, in consumption analysis it is the parameters of the utility function. The intuition behind the idea that technical change is local is that the knowledge accumulated through learning and investment is specific to those techniques that are in use, not to the complete spectrum of techniques that could be adopted ex ante. On the other hand, specific technological knowledge acquired through local learning cannot be easily codified and transferred, too. There are different ways of modelling the impact of experience on consumer behaviour. One may consider individuals who are initially nonconsumers of a good but become consumers thanks to the information acquired through consumption. Another strategy is based on the idea that consumption exposure generates experience and skills (e.g. mountain climbing) or that it allows one to discover one’s tastes. In this paper, we follow an intermediate strategy based on the idea that consumption generates both local information and experience that affect the level of effective satisfaction delivered by goods. Unfortunately, available models of cultural consumption with learning (e.g. Levy-Garboua & Montmarquette, 1996, 2002)5 cannot be adapted to our specific aims since they are based on the idea that there is a perfect information steady-state equilibrium whereas, in our model people never get to such a learning equilibrium. This outcome is due to the assumption made here that within the boundaries of our life, learning is an endless process. The latter assumption translates into the novel idea that, above a given threshold, learning does not depend on the quantity of a good consumed but on time spent consuming it. This rules out also work on habit formation (e.g. Stigler & Becker, 1977) positing that the stock of human capital generated is related to physical units of consumption. Instead, here we take learning as an unintentional process that takes time and argue that above a given threshold, the stock of consumption human capital is not directly related to the stock of consumption. Hence, the level of consumption cannot be strategically modified to optimize the learning process but, at best, to redistribute efficiently its gains.
Idiosyncratic Learning, Creative Consumption and Well-Being
47
Consumption exposure is assumed to activate three different feedback mechanisms from consumption to preferences and choices. The first works through hedonic and affective experience, i.e. the physiological and psychological forces that are stimulated by exposure to consumption; for example, the drinking of a glass of Barolo wine generates physiological and psychological stimuli that may affect consumption capital and the future perception and appreciation of Barolo (and of other red wines with similar characteristics).6 The second mechanism operates through the learning processes activated from consumption. Depending on the nature and complexity of a good (e.g. number and potential combinations of characteristics that generate hedonic stimuli), the capacity to enjoy consumption may require a training process based on exposure. Mountain climbing, skiing swimming, piano playing are clear examples. The third feedback mechanism concerns the generation of information about one’s preferences and, therefore, the expansion of the information set available to make choices. Owing to the local nature of learning, this information is also local. This assumption does not imply consumers’ myopia in the strong form assumed in the literature on habit formation (Pollak, 1970). Notably, we posit that, under certain conditions, individuals recognize the impact of current choices on future preferences, that is, they learn from their choices. The contention is that the opportunities for this learning process are restricted to an appropriate neighbourhood of the actual consumption choices, so that the unfamiliar choices are partially excluded from the learning process and do not generate information on preferences. Of course, the idea that familiarity with consumption affects choices is not new in the psychological literature and is consistent, in particular, with the idea that utility from consumption depends not only on direct experience but also on the memory of past and the anticipation of future consumption:7 ‘‘The impact of memory and anticipation on current utility leads to a type of triple counting of experience. A single event can influence utility first through anticipation, then through direct experience, and finally through memory’’ (Elster & Loewenstein, 1992, p. 214). The consumption of creative goods fits this idea very well in that it requires skills and the utility it delivers is significantly affected first by consumption exposure and then by the pattern of accumulation of consumption human capital. For instance, the appreciation of jazz, of an Impressionist painting or of mountain climbing are forms of creative consumption affected both by memory and by anticipation of past hedonic/affective experience. There are various ways to model how local consumption exposure affects consumption human capital8 and the evolution of preferences for creative
48
MARINA DI GIACINTO AND FRANCESCO FERRANTE
goods. In the following sections we develop some conjectures on the factors determining: (a1) the characteristics of the feedback mechanisms from consumption to preferences; (b1) the extension of the familiar consumption space within which exposure spills over and generates consumption human capital and information on the evolution of preferences.
3.1. The Characteristics of the Feedback Mechanisms The consumption of functional or comfort goods is determined by the traditional factors identified by microeconomics. Due to their nature as necessities, we assume that their consumption has constant returns over time and that the consumer’s spending on them can be treated as a deduction from income.9 The mechanisms leading to local feedbacks in creative consumption can be fruitfully set forth through the analysis of choices in the space of goods’ characteristics (Lancaster, 1971) by observing that the individuals are exposed only to the set, combinations and intensity of the characteristics that are present in the goods they consume. Leaving functional goods aside, let us assume that well-being is defined over consumption sets and that the latter are identified by groups of creative goods combining N characteristics in different manners (e.g. comfort, quality of design). Of course, one should expect N to be very large and consumption sets to differ not only in combinations of the same characteristics but also in the actual characteristics they present. In this scenario, consumers’ choices concern alternative creative consumption sets defined over a vector of characteristics that may contain some zeros. Hence, for simplification, we abstract from goods and suppose that in choosing consumers combine sets of characteristics in fixed proportions. This assumption preserves the idea that consumers can combine goods to obtain variety but posits that there is a limit to the process. Where standard consumer theory stresses the freedom to combine as a main ingredient of choice, we stress the idiosyncratic nature of choice as a process of selection among sets of alternatives. Let us assume that a consumption set m, m ¼ 1,2, y, M, is identified by a set of characteristics. The well-being delivered by consumption set m is assumed to be represented by the function:10 ðmÞ wðmÞ ¼ wðmÞ sðmÞ (1) t t ; xt
Idiosyncratic Learning, Creative Consumption and Well-Being
49
where: t, tZ0, is the time variable; s(m):[0,+N]-[0,N] is the stock of specific consumption human capital conditional on actual exposure at time t, which we call hedonic learning function; x(m):[0,+N]-[0,N] is consumption of set m. The function w(m) has to be twice differentiable and such that @wðmÞ =@sðmÞ ; @wðmÞ =@xðmÞ 40 and @2 wðmÞ =ð@sðmÞ Þ2 ; @2 wðmÞ =ð@xðmÞ Þ2 o0; i.e. strictly increasing and strictly concave in both variables s(m) and x(m). In particular, we argue that consumption human capital changes only for those creative goods that are objects of consumption above a minimum exposure threshold x¯ ðmÞ at least once, i.e. if there exists tðmÞ such that xðmÞ x¯ ðmÞ ; and that the pattern of accumulation of consumption human ¯tðmÞ capital is specific to an individual and a set. As far as creativity is concerned, we posit that the objective creativity content of a given set is increasing in its complexity, the latter being given by the number of characteristics and by their potential combinations.11 A reasonable representation of the evolution of s(m) builds on the idea of a life-cycle pattern in the consumption. Early, until a time tðmÞ max 40; the exploration of new creative consumption opportunities generates increasing ðmÞ returns over time, i.e. DsðmÞ t =Dt 0; 8totmax : Individuals learn how to enjoy the stimuli generated by the specific combinations of characteristics provided by a given set. These consumption skills improve over time as individuals learn to explore creative consumption opportunities. At a later stage, owing to the acquisition of skills and the depletion of the stock of novelty, repetition of consumption drives returns from positive to negative ðmÞ values, i.e. DsðmÞ t =Dt 0; 8t4tmax : (m) The shape of the curve s that describes the life-cycle of specific consumption human capital depends both on objective characteristics, of the set m, i.e. how creative a good is in terms of number and complexity of its characteristics, and on subjective characteristics of individuals, i.e. their aptitude to exploit the creative potential of the consumption set. Therefore, objective properties of goods and subjective characteristics of individuals interact in determining the actual content of creativity and the evolution of preferences through the accumulation of consumption human capital. A very good example is jigsaw puzzles. Appreciation of this activity depends on an individual’s innate taste and ability to do puzzles, on ability to do puzzles acquired through experience and, finally, on the objective difficulty of the puzzle (number of pieces and complexity of the pattern).
50
MARINA DI GIACINTO AND FRANCESCO FERRANTE
Indeed, individuals show different attitudes towards newness in consumption, aptitude to learn from exposure. Some may love newness as such, and then become quickly bored; others may be less concerned for novelty and less bored by repetition. Moreover, the opportunities to learn from exposure may vary greatly from one individual to another depending on how actively they are involved in consumption. Going to a pop concert or to opera could be a very creative activity, yielding high returns for some individuals, and a pretty boring one for others. Building on our arguments, we suppose that the feedback from consumption to preferences may be positive or negative and lead either to accumulation or depletion of human capital depending on: (a2) the specific stage within the life-cycle of a given consumption set; (b2) an individual’s type; (c2) the nature of the creative set. A suitable mathematical representation of the hedonic learning function is 8 gðmÞ ttðmÞ > ðmÞ < sðmÞ þ ttðmÞ e bðmÞ if 9tðmÞ such that xðmÞ ðmÞ x 0 2 t sðmÞ (2) :¼ t > : sðmÞ otherwise 0
where:
ðmÞ sðmÞ 0 ¼ s0 ðI; KÞ is a function concerning set m that depends on innate tasteI and on human capitalK acquired without exposure to consumption. We assume that it is twice differentiable in both variable; ¯t ðmÞ is the arrival time of set m, i.e. the first time, if it exists, such that xðmÞ ¯ ðmÞ ; ¯t ðmÞ x ðmÞ ðmÞ ðmÞ the quantity ðt ¯tðmÞ =2Þg et¯t =b describes at any time the accumulation of consumption human capital; the parameter g(m), 1rg(m)o+N, represents the objective creativity of consumption set m. In particular, g(m) ¼ 1 is the value of ordinary functional goods, and the higher the value of the parameter, the greater its creativity (see point (c2) above); the parameter b(m), 1rb(m)o+N, describes the subjective creativity concerning the consumption of set m; the higher subjective creativity, the larger the value of b(m) (see point (b2) above).
A life-cycle is defined over the time interval [0, T(m)], where t ¼ 0 is the starting period of consumption and t ¼ T(m), T(m)>0, is such that sðmÞ ¼ T ðmÞ sðmÞ þ ; where e>0 is arbitrarily small. 0
Idiosyncratic Learning, Creative Consumption and Well-Being
51
Hence, with zero exposure, the amount of well-being delivered by set m depends on s0ðmÞ : So, as a matter of analytical simplicity, we assume that information about appreciation without exposure is unbiased. From (2) one can verify that as soon as demand for a consumption set reaches the minimum exposure threshold, which is not known in advance to consumer, learning no longer depends on consumption level. We can notice that the larger g(m) or b(m) is the higher the value of the hedonic learning function s(m). Moreover one can easily verify that s(m) ðmÞ ðmÞ reaches its absolute peak at tðmÞ b þ ¯tðmÞ with maximum value max ¼ g ðmÞ ðmÞ ðmÞ ðmÞ ðmÞ s ðmÞ ¼ s0 þ ðg b =2eÞg : Both are increasing in parameters g(m) and b(m), tmax so the time required for the consumer to reach maximum hedonic exposure increases with objective and subjective creativity, and the global maximum value of the hedonic learning also increases. One can draw different shapes of s(m), over an individual lifetime, on the basis of the tastes of the individual and the characteristics of the set, i.e. g(m) and b(m). Figs. 1a and 1b show three hedonic functions with different shapes.12 In Fig. 1b we test the sensitivity of the hedonic learning function with respect to objective and subjective creativity.13 Other things being equal, it is more sensitive to changes in objective than in subjective creativity. The area S(m) under curve s(m) can be adopted as a reasonable measure of the potential creativity of consumption set m: S ðmÞ :¼ T ðmÞ
T ðmÞ
Z ¯t ðmÞ sðmÞ dt ¼¼ sðmÞ t 0 dt 0 0 # gðmÞ Z T ðmÞ " ¯t ðmÞ t ¯t ðmÞ tðmÞ ðmÞ þ dt e b s0 þ 2 ¯tðmÞ
Z
ð3Þ
since it is an increasing function of the parameters g(m) and b(m). Solving the previous integral we obtain: ðmÞ S ðmÞ ¼ sðmÞ þ 0 T T ðmÞ
ðbðmÞ Þg 2
ðmÞ
gðmÞ
þ1
G¯tðmÞ ;T ðmÞ ðgðmÞ þ 1Þ
(4)
where G¯t ðmÞ ;T ðmÞ ðgðmÞ þ 1Þ is the Euler generalized incomplete gamma function.14
52
MARINA DI GIACINTO AND FRANCESCO FERRANTE
Fig. 1. (a) Hedonic functions with different levels of objective and subjective creativity. (b) Sensitivity of the hedonic learning functions with respect to objective and subjective creativity.
In the construction of the consumption space without exposure, building on our notion of creative consumption and idiosyncratic learning, we posit that the following conditions are satisfied: (a3) for a given level of creativity, the productivity of exposure in generating ðmÞ initial appreciation sðmÞ 0 is increasing in human capital, i.e. @s0 =@K40;
Idiosyncratic Learning, Creative Consumption and Well-Being
53
(b3) given innate preferences, appreciation without exposure sðmÞ depends 0 on how familiar the characteristics of a set are with respect to the nature of one’s human capital: if set n is more familiar than set m, then ZsðnÞ ðKÞ4ZsðmÞ ðKÞ; where ZsðnÞ ðKÞ and ZsðmÞ ðKÞ are respectively the elas0 0 0 0 ðmÞ ticity of sðnÞ with respect to variable K, i.e. ZsðnÞ ðKÞ :¼ 0 and s0 0 ðnÞ ðnÞ ðmÞ @s0 =@K K=s0 and ZsðmÞ ðKÞ :¼ @s0 =@K K=sðmÞ 0 ; 0 (c3) K affects the appreciation of more creative consumption goods more strongly, i.e. if set n is more creative than set m, then ZsðnÞ ðKÞ4ZsðmÞ ðKÞ; 0 0 (d3) given innate preferences, appreciation without exposure sðmÞ depends 0 on the specificity of human capital; the more specific K, the more (less) effective it is in affecting the appreciation of more (less) familiar consumption sets. The previous conditions seem quite reasonable. The appreciation of any creative activity without exposure depends on the amount of human capital we possess, and in particular on the amount of specific skills we can deploy. For instance, the appreciation of mountain climbing without experience depends on our human capital, and specifically on that necessary to enjoy climbing. If we have practiced mountain hiking before, we will appreciate climbing more rather than less. On the other hand, the appreciation of jazz depends more heavily on human capital than does the appreciation of pop music, as require certain skills. Of course, this does not imply that listening to jazz would provide more well-being than listening to pop music for any given individual.
3.2. Consumption Choices with No Exposure We suppose that the choice ab ovo among consumption sets is determined through an optimization process based on full knowledge of consumption appreciation sðmÞ without exposure but no information on the future evo0 lution of preferences. This process results in choosing the level of consumption of different sets, xðmÞ at time t ¼ 0; that maximises unconditional 0 ð2Þ ðMÞ ð1Þ ð2Þ ðMÞ well-being W 0 ¼ W ðsð1Þ 0 ; s0 ; . . . ; s0 ; x0 ; x0 ; . . . ; x0 Þ defined as follows: W 0 :¼
M X m¼1
wðmÞ 0 ¼
M X
ðmÞ wðmÞ ðsðmÞ 0 ; x0 Þ
(5)
m¼1
i.e. well-being without consumption exposure delivered by M available consumption sets, given their respective prices15 p(1), p(2), y, p(M) and the exogenous income Yc devoted to creative consumption. Given innate taste
54
MARINA DI GIACINTO AND FRANCESCO FERRANTE
and human capital, an individual selects consumption x0nðmÞ that maximizes well-being, i.e. max
xð1Þ ;xð2Þ ;...;x0ðMÞ 0 0
W0 ¼
M X
max
xð1Þ ;xð2Þ ;...;xðMÞ m¼1 0 0 0
s:t:
ðmÞ wðmÞ ðsðmÞ 0 ; x0 Þ
8 ð1Þ ð2Þ ðMÞ < x0 pð1Þ þ xð2Þ þ þ xðMÞ ¼ Yc 0 p 0 p : xðmÞ 0; m ¼ 1; 2; . . . ; M 0
ð6Þ
ðmÞ ðmÞ a Let us assume wðmÞ 0 ¼ ½s0 x0 ; 0oao1, and M ¼ 2. Then, solving the optimization problem (6) at time t ¼ 0, we obtain the following standard demand functions: a=ð1aÞ pð2Þ sð1Þ 0 x0ð1Þ ¼ Y c (7a) a=ð1aÞ a=ð1aÞ ð2Þ sð1Þ pð1Þ pð1Þ sð2Þ þ p 0 0
x0ð2Þ ¼ Y c
pð2Þ
pð1Þ sð2Þ 0
pð1Þ sð2Þ 0 a=ð1aÞ
a=ð1aÞ
a=ð1aÞ þ pð2Þ sð1Þ 0
(7b)
Here we can draw several insights into the expected behaviour in the initial stage of the individual life-cycle in the purchase of objective creativity.16 First of all, we observe that hedonic learning is constrained by income: given the solution of the optimization problem (6), it can be easily verified that learning through consumption of set 1 and set 2 is allowed if and only if respectively: 2 3 ! a ð1Þ ð2Þ ð1aÞ p s 0 Y c x¯ ð1Þ pð1Þ 4 þ 15 (8a) pð2Þ sð1Þ 0 and
2
pð2Þ sð1Þ 0 Y c x¯ ð2Þ pð2Þ 4 pð1Þ sð2Þ 0
a !ð1aÞ
3
þ 15
(8b)
Now rank consumption sets by increasing objective creativity, so that if n4m than set n is more creative than set m. It is reasonable to hold that,
Idiosyncratic Learning, Creative Consumption and Well-Being
55
ceteris paribus, more creative sets must be consumed in larger quantities to generate hedonic learning, i.e. the minimum exposure thresholds are such ðmÞ that x¯ ðnÞ 4x¯ ðmÞ : Furthermore we consider reasonable that sðnÞ and 0 s0 ðnÞ ðmÞ p p ; i.e. the value of the initial appreciation without exposure and the price of the more creative set n are respectively lower and higher than those ð1Þ of set m. In the case of M ¼ 2, even in the borderline case in which sð2Þ 0 ¼ s0 (2) (1) and p ¼ p , it can straight verified from (8a) and (8b) that the consumption set with greater objective creativity, set 2, requires more income to generate hedonic learning. Second, due to the essential contribution of human capital to the appreciation of the more creative consumption sets, ceteris paribus,17 individuals with inadequate K will be relatively more stringently rationed in purchasing of objective creativity and, thereby, a reduction (increase) in human capital K involves a reduction (increase) in demand for the more creative set 2 and an increase (a reduction) in demand for set 1. As a matter of fact it can be proved that18 @x0nð2Þ =@K40; and consequently @x0ð1Þ =@Ko0; if and only if ZSð2Þ ðKÞ4ZSð1Þ ðKÞ; and the latter inequality always holds if set 2 is more 0 0 creative then set 1 by assumption (c3) in Section 3.1. Hence, the more creative sets are likelier to be excluded from hedonic learning opportunities by income and human capital constraints.
3.3. The Extension of the Learning Region and Idiosyncratic Accumulation of Human Capital Building on our previous assumptions and with income exogenous, subsequent choices depend on the endogenous change in appreciation of the available consumption sets. This change stems from consumption exposure and the cultivation of taste. Let us define at time tZ0, the set of learninggenerating sets Lt in the following way: n o ðnÞ Lt :¼ n 2 Mj9¯t ðnÞ 0jxn¯t ðnÞ x¯ ðnÞ ; ¯t ðnÞ t (9)
where Mis the set of all available consumption sets, i.e. M ¼ f1; 2; . . . ; M g; C and indicate by LC t the residual set, i.e. Lt :¼ M Lt : Taste can be cultivated either directly, for sets consumed above the learning threshold, or indirectly, for sets that, though consumed below that threshold, enjoy the spillovers generated by the other. We would expect different individuals and consumption sets to be characterized by varying opportunities for spillovers. For example, one would
56
MARINA DI GIACINTO AND FRANCESCO FERRANTE
expect large spillovers between different forms of visual art consumption or sports requiring the same skills but virtually none between the two groups. Let us assume that at time tZ0, the creative set m is consumed below x¯ ðmÞ ; i.e. m 2 LC t ; and that there exists at least one consumption set n consumed above x¯ ðnÞ ; i.e. nALt. Building on the previous formulation, we posit that the appreciation at time t of set m, in this case, is represented by the following function: gðmÞ ðmÞ t tðmÞ tt ðmÞ ðmÞ st :¼ s0 þ (10) e bðmÞ max jðm;nÞ ðyÞ n2Lt 2
where:
y:C C-[0, +N) is a suitable measure of familiarity, i.e. the distance among consumption sets in the space of characteristics C;19 j(m,n): [0, +N)-[0, 1] is the spillover function describing the transferability of hedonic learning from consumption set n to set m, which we posit as monotonically decreasing; moreover, if y ¼ 0 than j(m,n)(0) ¼ 1 and spillovers are complete, and when y-+N than lim jðn;mÞ ðyÞ ¼ 0 y!þ1 and there are n no spillovers; o tðmÞ :¼ min t¯ðnÞ 0jjðm;nÞ 40 , i.e. t(m) is the first point in time at which the accumulation of consumption human capital begins to increase either indirectly via spillover or directly because set m begins to be consumed above the minimum exposure threshold x¯ ðmÞ :
We remark here that if there exists more than one consumption set, at time tZ0, consumed above the minimum exposure threshold, spillovers are generated by the closest set n to set m within the set of all available consumption sets M: Therefore, in the presence of spillovers, the hedonic learning function, can be described as follows: 20 8 ðmÞ > if Lt ¼ + < s0 ; ðmÞ gðmÞ ttðmÞ (11) sðmÞ :¼ t ðmÞ tt > e bðmÞ maxfjðm;nÞ ðyÞg; if Lt 6¼ + : s0 þ 2 n2L t
We notice that the function in (11) is a more general case of the function defined in (2). As a matter of fact when tðmÞ ¼ ¯tðmÞ ; since n ¼ m, then y ¼ 0 and j(m,m)(0) ¼ 1. The intuition that learning from consumption exposure is local and that the accumulation of consumption human capital is specific finds support – e.g. the consumption of music – in the empirical evidence provided by
Idiosyncratic Learning, Creative Consumption and Well-Being
57
Kurabayashi and Ito (1992) which show a positive correlation of demand between different types of music within genres and a negative one between genres. The reasoning behind the idea that the distance in the space of characteristics is a good proxy for familiarity in consumption is that, through the accumulation of consumption human capital, individuals get to know the specific set and the combinations of characteristics of the goods they consume most; and that taste acquired through cultivation can be transferred only to those goods with similar sets and combinations of characteristics. An important implication of this is that, at least in the case of most creative goods, for preference formation information is a highly imperfect substitute for personal consumption exposure (on this point, see Ferrante, 1999). Note that the idea of diminishing consumption spillovers, as portrayed here, has its counterpart in the property of diminishing sensitivity discussed by Kahneman and Tversky in their analysis of reference-dependent preferences: ‘‘The impact of a difference [between two consumption bundles, measured on a given dimension] is attenuated when both options are remote from the reference point for the relevant dimension’’ (Tversky & Kahneman, 1990, p. 1040). An interesting question is whether spillovers should be assumed symmetric, i.e. j(m,n) ¼ j(n,m), or if, instead, they should be assumed to depend on the characteristics of sets and, specifically, on creativity. We contend that more creative sets generate larger spillovers, e.g. if set n is more creative than m, then j(m,n)4j(n,m). A suitable representation of a spillover function, embodying the previous properties, is the following: jðm;nÞ ðyÞ :¼
1 ð1 þ lðm;nÞ Þy
(12)
where l(m,n), 0rl(m,n)r1, is a parameter measuring the capacity of consumption set n to generate spillovers in favour of set m. Building on the previous arguments, l(m,n) is assumed to relate negatively to human capital and to the creativity of n, and positively to the specificity of human capital with respect to the characteristics of n.
3.4. Local Learning and Inter-Temporal Choices It is now possible to predict the evolution of preferences conditional on initial choices and the associated hedonic learning regimes. Determining the
58
MARINA DI GIACINTO AND FRANCESCO FERRANTE
characteristics of the initial process of selection is essential in a world where consumption history matters.21 It is worth repeating again that in our model freedom of choice does not imply that people can affect the learning process by redistributing consumption over time. Given the initial consumption choices, learning is generated autonomously above a given consumption threshold. The evolution of preferences is determined by two basic factors, the shape of the hedonic learning function of each consumption set and, in addition, the shape of the spillover function associated with each set belonging to L. In short, the information needed to generate and investigate the inter-temporal decision-making scenario is the set of functions sðmÞ ; 8m 2 L; and the set of spillover functions jðm;nÞ ; 8m 2 L and 8n 2 LC : Idiosyncratic learning also concerns the private information that is generated by consumption exposure and used to re-optimize the decision process over time. The latter information is necessary to up-date preferences. If one assumes local perfect foresight,22 i.e. ðmÞ E½sðmÞ ¼ sðmÞ t þ t ; 8t 0
(13)
it is easy to show that the maximization of well-being over the entire life requires marginal well-being to be equated across time. Hence, consumption must be redistributed over time to allocate efficiently23 the expected gains from creative consumption generated by the set of functions sðmÞ ; 8m 2 L: As one would expect, local perfect forecast yields locally optimal consumption profiles that are also globally optimal if one considers as given, income, prices and the ex ante endowment of consumption human capital. Unfortunately, empirical evidence about people’s ability to predict future preferences is quite discouraging for the supporters of perfect foresight and endures some sort of myopia (Loewenstein & Adler, 1995). This is not surprising if one considers that the object of cognitive learning is not an external entity such as, for example, the price of cars, but the physiological and psychological mechanisms internal to individuals. Experimental evidence on drugs consumption provides even stronger results. Individuals may take instantaneous decisions affecting their consumption history that are driven by their emotional sphere and that contrast with their rational perception of what is good for them (Bernheim & Rangel, 2004). Here we assume that individuals have a perfect instantaneous perception of the building-up of such mechanisms but that they are unable to make forecast about the future evolution of the latter. This assumption allows us to adopt a static analytical setting.
Idiosyncratic Learning, Creative Consumption and Well-Being
59
With M ¼ 2, and assuming that consumption set 1 belongs to Lt and consumption set 2 belongs to LC t ; well-being at time tZ0, is given by W t ¼ ð1Þ ð1Þ ð1Þ ð2Þ ð2Þ ð2Þ w ðst ; xt Þ þ w ðst ; xt Þ: With zero time discount rate, the problem to solve is: h i ð1Þ ð2Þ ð1Þ ð2Þ max W t ¼ max wð1Þ sð1Þ ; x s ; x þ w t t t t ð2Þ xð1Þ t ; xt
ð2Þ xð1Þ t ; xt
s:t:
8 ð2Þ ðMÞ < xtð1Þ pð1Þ þ xð2Þ þ þ xð2Þ ¼ Yc t p t p : xð1Þ ; xð2Þ 0 t t
ð14Þ
ð1Þ a ð2Þ ð2Þ a 24 For W t ¼ ðsð1Þ t xt Þ þ ðst xt Þ ; 0rar1 , at time tZ0, one gets: a ð1aÞ pð2Þ sð1Þ t (15a) xtnð1Þ ¼ Y c a a ð2Þ ð1aÞ ð1Þ ð1aÞ ð1Þ ð1Þ ð2Þ p p st þ p st
xtnð2Þ
¼
a ð1aÞ pð1Þ sð2Þ t Yc a a ð2Þ ð1aÞ ð1Þ ð1aÞ ð2Þ ð1Þ ð2Þ p p st þ p st
(15b)
Of course, the solution to the static optimization problem does not correspond to the solution of the dynamic one. But, if one considers our assumptions as a more plausible story, i.e. a story more consistent with experimental evidence, the resulting equilibrium behaviour is not deprived of optimality properties. Indeed, it corresponds to the best one can do given the assumptions about the nature of information underlying idiosyncratic learning. In the general case, a non-trivial scenario is one where, at time t ¼ 0, at least one set is consumed below and at least another one above the learninggenerating threshold and spillovers are not complete, i.e. the spillover function assumes positive values less than 1. In view of the characteristics of individuals and sets and their impact on the shape of, respectively, the hedonic and the spillover functions, the number of possible scenarios is large. The question one may wish to address is how the endogenous change in taste and other, exogenous, changes affect the opportunities for creative consumption and well-being over time. Several potential sources of change deserve to be investigated: (a5) endogenous changes, brought about by the simple passage of time, involving a pre-existing set of creative goods;
60
MARINA DI GIACINTO AND FRANCESCO FERRANTE
(b5) exogenous change due to income or relative price changes; (c5) exogenous consumption innovations involving newly introduced creative goods that expand the opportunities to generate new consumption sets. Given the specific aim of this paper, we leave out (c5) and discuss (a5) and (b5).25 As far as endogenous changes are concerned, the consumption of creative sets and the corresponding well-being change through time according to the endogenous evolution of preferences determined by the hedonic functions. Individuals enjoy the fruits of hedonic learning and exploit differences in the rate of accumulation and depletion of consumption human capital by substituting one set for another. Such substitution is governed by changes in the marginal rates of substitution between sets within the spillover region. Demand for and well-being delivered by a set increase when consumption human capital accumulates and vice versa. Demand and well-being may show jumps when the scenario described either at point (a4) or (b4) of footnote 20 occurs, and individuals suddenly start to enjoy the fruits of creative consumption. Hence, well-being can follow different patterns depending on (i) the rates at which consumption human capital accumulates and depletes for different sets and individuals and (ii) the actual generation of spillovers. The resulting shape of the hedonic functions determines the shape of the well-being function, which may be either monotonically or non-monotonically increasing with respect to time. Notably, well-being may be steadily increasing, or it may have several peaks depending on the distribution of the gains from the accumulation of consumption human capital over time.26 Given subjective creativity, total life well-being is determined by the total amount of objective creativity available and accessible through the life-cycle to an individual. Income changes bring about both standard and non-standard demand effects. The standard effect stems for the positive relation between demand and income: when income rises, so does demand for all goods. The nonstandard effect occurs when, the increase in demand for set n, previously consumed below the learning threshold, activates hedonic learning for that set. In this case, as income rises, demand for set m may either increase or decrease, depending on how strong the hedonic learning effect is.27 Changing prices may lead to the same threshold effect. What about the link between income and well-being over time? Of course, at each point in time, income has a positive effect on well-being in that it increases consumption and can activate hedonic learning. Conversely, over an individual life-cycle, constant income growth could prove insufficient to
Idiosyncratic Learning, Creative Consumption and Well-Being
61
sustain non-decreasing well-being if the rates of depletion of consumption human capital are relatively high. When income rises at a constant rate, the condition for non-decreasing well-being can be easily derived by calculating the time derivative of the well-being function. Of course, one should expect that decreasing well-being in the presence of increasing income will more likely occur at the later stages in individual life and in scarcely innovative environments. Whenever income and price changes activate hedonic learning mechanisms sufficiently stronger than those experienced in the past, individuals may feel a particular form of regret28 for previous choices and reverse her/his revealed preference. In fact, they may realize that more creative and feasible creative consumption sets were available in the past, i.e. consumption sets yielding higher well being – even if valued at old income and relative price – than those actually experienced. A large body of empirical evidence in different fields indicates that the satisfaction people derive from their experiences is influenced by their perception of what would have occurred had they made different choices. Even if our action was optimal given the information available ex ante, we nevertheless experience regret when we later discover that some other action would have made us better off (Krahmer & Stone, 2005). We assume that whenever regret occurs, well-being is reduced by an amount Wr positively related to the perceived loss of well being.29 The consumer can be likened to a mountain climber who, on her/his way towards the peak, surveys the landscape below and realizes that she/he could have chosen an easier way. Of course, owing to the informative constraints binding individuals, nor regret neither preference reversal imply inefficient or inconsistent inter-temporal choices. The model and the previous discussion provide useful insights about the characteristics of individuals and goods that should be factored in empirical analysis of creative consumption. Although an articulated discussion of the implications of the model for empirical analysis is behind the scope of this paper, nevertheless, it is worth emphasising here that the type of education of consumers, their age and their parents’ education, the type of good and their degree of objective novelty are important elements to be taken into account.
4. SUMMARY AND CONCLUSIONS The model of idiosyncratic learning developed in this paper does not pretend to give a full account of the factors affecting consumers’ inter-temporal
62
MARINA DI GIACINTO AND FRANCESCO FERRANTE
choices. Its far more limited aim is to offer a coherent description of the effects of assuming that the accumulation of consumption human capital is an idiosyncratic process and that hedonic and affective experiences cannot be codified and transferred. The model provides useful insights about important factors that should be factored in empirical analysis of creative consumption, such as the level and type of education of consumers, their age and their parents’ education. The theoretical implications of this simple model are quite disturbing for standard economic doctrine as short-run choices governed by innate taste may well lead one under-develop preferences requiring high consumption skills and exposure, which are needed to appreciate the consumption of most creative goods. We show that idiosyncratic learning may well imply that individuals experience dynamic regret for previous choices and preferences reversal but that the latter outcomes do not imply inefficient or inconsistent choices. Given the nature of the physiological and psychological mechanisms governing the evolution of preferences, individuals could not do better than they do. On the other end, it is shown that individuals may be rationed in purchasing creative goods – those that can deliver greater lifetime well-being – mostly due to human capital and income constraints. Owing to the essential role played ab ovo by education and exposure to consumption in the formation of preferences, the model suggests that appropriate education policies may improve cognitive and learning abilities of people and their potential for creative consumption.
NOTES 1. The expressions consumption styles and consumption sets are used interchangeably. 2. Loss aversion and diminishing sensitivity are two main features of the model of Tversky and Kahneman that one can find in the model presented here. 3. The idea that personal consumption experience can be codified and de-codified in a very imperfect way is an implicit assumption that social interaction is a poor device to share information about consumption opportunities. 4. Psychobiology lends strong support to the idea that learning processes generated by consumption experience is idiosyncratic and cannot be codified. Emotional and affective mechanisms activating consumption learning work through chemical transformations of the cerebral cells and the nature of this process is specific to individuals and to the type of consumption experiences.
Idiosyncratic Learning, Creative Consumption and Well-Being
63
5. They develop a model where the equilibrium is achieved when the subjective qualities of the goods have stabilized at the path-dependent stationary values determined at the end of the learning period. 6. For example, structure, taste, appearance and bouquet. 7. ‘‘The role played by past consumption experience in determining future choices is crucial in that unlike the future, the past cannot be altered, so that its effects on the present are largely determined by prior decision. Our current selves are largely at the mercy of past selves, although we have some limited capacity to direct our thoughts towards or away from the past or even to represent the past as we wish. But if past selves have endowed us with an overly rich or lean stock of memories, there are limits to our capacity to cognitively amend them without lapsing into autism’’ (Elster & Loewenstein, 1992, p. 214). 8. Indeed, in the context of creative goods, the traditional distinction between cultural and consumption human capital tends to disappear and the two concepts can be used interchangeably. 9. One can reasonably assume that the share of such expenditure is decreasing in income. 10. We believe that, when dealing with learning, ‘‘utility’’ should be replaced by ‘‘well-being’’ or ‘‘happiness’’ (see Frey & Stutzer, 1999). 11. Of course, combinations should be meaningful to the individual, i.e. they should provide positive stimuli. ð2Þ ð3Þ ¯ð1Þ ¼ ¯tð2Þ ¼ 0 and ¯tð3Þ ¼ 12. In Fig. 1a we assume that sð1Þ 0 ¼ s0 ¼ s0 ¼ 10 while t 9: Moreover, we consider different values for objective and subjective creativity but we impose that gð1Þ bð1Þ ¼ gð2Þ bð2Þ ¼ gð3Þ bð3Þ ¼ 30; hence the global maxima are re(1) ð2Þ ð3Þ spectively at tð1Þ ¼ 2.8, g(2) ¼ 3, max ¼ tmax ¼ 30 and tmax ¼ 39: In particular, we set g (3) (1) (2) (3) g ¼ 3.3 and therefore b ¼ 75/7, b ¼ 10, b ¼ 100/11. ð2Þ ð3Þ ¯ð1Þ ¼ ¯tð2Þ ¼ ¯tð3Þ ¼ 0: The function s(1) has the 13. We set sð1Þ 0 ¼ s0 ¼ s0 ¼ 10 and t same values of objective and subjective creativity g(1) ¼ b(1) ¼ 4.1. In s(2) the parameter g(2) increases of 10%, i.e. g(2) ¼ 4.51 and b(2) ¼ 4.1. Finally in function s(3) the objective creativity g(3) is unchanged respect to that of s(1) while the subjective creativity b(3) is augmented of 10%, i.e. g(3) ¼ 4.1 and b(3) ¼ 4.51. 14. The Euler generalized incomplete gamma function is given by G¯tðmÞ ;T ðmÞ R T ðmÞ ðmÞ ðgðmÞ þ 1Þ :¼ ¯tðmÞ zg ez dz; that we immediately obtain setting z ¼ t tðmÞ =bðmÞ ; R þ1 ðmÞ while the Euler complete gamma function satisfies GðgðmÞ þ 1Þ :¼ 0 zg ez dz and R þ1 ðmÞ the Euler incomplete gamma function satisfies GT ðmÞ ðgðmÞ þ 1Þ :¼ T ðmÞ zg ez dz: 15. The price of a set is a composite price of the goods embodied in the set. 16. Of course, the two-set case can be easily generalized. 17. That is, positing equal familiarity, with respect to human capital, of set 1 and set 2 (see point (b3) in Section 3.1). ð2Þ ð2Þ ð2Þ ð2Þ ð1Þ ð1Þ 18. @xð2Þ and @xð1Þ 0 =@K ¼ @x0 =@s0 @s0 =@K þ @x0 =@s0 @s0 =@K 0 =@K ¼ ð1Þ ð1Þ ð1Þ ð1Þ ð2Þ ð2Þ @x0 =@s0 @s0 =@K þ @x0 =@s0 @s0 =@K: 19. For instance, one may adopt the Euclidian distance in the space of characteristics C or a normalised measure of the distance between C(m) and C(n). 20. We observe that this hedonic learning function can have a finite number of discontinuity points (no more than M1) which, according to our economic
64
MARINA DI GIACINTO AND FRANCESCO FERRANTE
discussion, represent jumps of the hedonic learning function that can be generated in two different scenarios: (a4) when set m is consumed below the minimum exposure threshold x¯ ðmÞ and (i) at the same time there is at least one set n consumed above the minimum threshold that generates incomplete spillovers, (ii) there exists a time ¯t ðkÞ 4¯t ðnÞ ; such that the consumer starts to consume another set k above the exposure threshold x¯ ðkÞ and the latter generates more spillovers than set n; (b4) when set m is consumed below the minimum exposure threshold x¯ ðmÞ and (i) at the same time there is at least one set n consumed above the minimum threshold that generates incomplete spillovers, (ii) there exists a time ¯t ðmÞ 4¯t ðnÞ ; such that consumption x(m) reaches the threshold x¯ ðmÞ At these points of discontinuity, we conventionally assume that @sðmÞ =@t ¼ þ1: 21. The act of choosing has a positive component, the selection of something, but also a negative component, the exclusion of everything else! The consequences of the exclusion are, generally speaking, neglected in economics. 22. Where E[ ] is the expectation operator and the random variable A(m) represents the forecast error concerning the consumption set m; moreover A(m) 2 ðmÞ N(0, s ) and are i.i.d. m2M 23. Of course, in the absence of liquidity constraints. 24. Demand functions x*(1) and x*(2) could have a finite number of discontinuity points corresponding to the jumps in the hedonic learning functions. We conventionally assume that at those points @xnð1Þ =@t ¼ þ1 and @xð2Þ =@t ¼ 1 if the jumps concern function s(1), while @xnð1Þ =@t ¼ 1 and @xnð2Þ =@t ¼ þ1 if the jumps concern function s(2). 25. Numerical examples applied to our model are provided in Appendix A. 26. The latter scenarios are shown, in the two good case, in Figs. A4a and A4b of Appendix B, while in Figs. A5a and A5b of Appendix B we depict the corresponding hedonic learning functions. 27. An example is given in Appendix C. 28. In this certainty context ‘‘regret’’ is used in a non-technical way. 29. It is worth emphasising here that the availability of a superior consumption life-style is not due to the new income/set of prices: the new income/set of prices allow to the consumers to recognizing this lost opportunity.
ACKNOWLEDGMENT We would like to acknowledge valuable comments on a previous version of this paper from Marina Bianchi, Bruno S. Frey, Maurizio Pugno and participants at various seminars. The usual disclaimer applies. Financial support from Ministero dell’Universita` e della Ricerca Scientifica e Tecnologica is gratefully acknowledged.
Idiosyncratic Learning, Creative Consumption and Well-Being
65
REFERENCES Atkinson, & Stiglitz, J. (1969). A new view of technical change. Economic Journal, 79, 573–578. Becker, G. S., & Murphy, K. M. (1988). A theory of rational addiction. Journal of Political Economy, 96, 675–700. Bernheim, D., & Rangel, A. (2004). Addiction and cue-triggered decision processes. American Economic Review, 94(5), 1558–1591. Elster, J., & Loewenstein, G. (1992). Utility from memory and anticipation. In: G. Loewenstein & J. Elster (Eds), Choice over time (pp. 213–234). New York: Russell Sage Foundation. Ferrante, F. (1999). Locally rational affection and the choice of environmental policy instruments. In: T. Sterner & T. Tietenberg (Eds), The market and the environment. The effectiveness of market based policy instruments for environmental reform (pp. 66–91). Cheltenham, U.K. and Northampton, MA: Edward Elgar. Frey, B. S., & Stutzer, A. (1999). Maximizing well-being? Working Paper no. 22, Institute for Empirical Research in Economics, University of Zurich. Kahneman, D., & Snell, J. (1992). Predicting a changing taste: Do people know what they will like?. Journal of Behavioral Decision Making, 5, 187–200. Krahmer, D., & Stone, R. (2005). Dynamic regret theory. Unpublished Working Paper. Kurabayashi, Y., & Ito, T. (1992). Socio-economic characteristics of audiences for Western classical music in Japan. In: R. Towse & A. Khakee (Eds), Cultural economics (pp. 275–287). Berlin: Springer-Verlag. Lancaster, K. J. (1971). Consumer demand: A new approach. New York: Columbia University Press. Layard, R. (2003). Happiness: Has social science a clue? Lionel Robbins Memorial Lecture, London School of Economics, March. Levy-Garboua, L., & Montmarquette, C. (1996). A microeconomic study of theatre demand. Journal of Cultural Economics, 20, 25–50. Levy-Garboua, L., & Montmarquette, C. (2002). The demand for the arts. Working Paper, CIRANO. Loewenstein, G. (1988). Frames of mind in inter-temporal choice. Management Science, 34(2), 200–214. Loewenstein, G. (1992). The fall and rise of psychological explanations in the economics of inter-temporal choice. In: G. Loewenstein & J. Elster (Eds), Choice over time (pp. 3–34). New York: Russell Sage Foundation. Loewenstein, G., & Adler, D. (1995). A bias in the prediction of tastes. Economic Journal, 105, 929–937. Loewenstein, G., & Thaler, R. H. (1989). Anomalies. Inter-temporal choice. Journal of Economic Perspectives, 3(4), 181–193. Pollak, R. A. (1970). Habit formation and dynamic demand functions. Journal of Political Economy, 78, 60–78. Pollak, R. A. (1976). Habit formation and long-run utility functions. Journal of Economic Theory, 13, 272–297. Pollak, R. A. (1978). Endogenous tastes and in demand and welfare analysis. American Economic Review, 68, 374–379. Scitovsky, T. (1992). The joyless economy. The psychology of human satisfaction. New York: Oxford University Press.
66
MARINA DI GIACINTO AND FRANCESCO FERRANTE
Stigler, G. J., & Becker, G. S. (1977). De Gustibus Non Est Disputandum. American Economic Review, 67(2), 76–90. Thaler, R. (1980). Toward a positive theory of consumer choice. Journal of Economic Behavior and Organization, 1, 39–60. Tversky, A., & Kahneman, D. (1986). Rational choice and the framing of decision. Journal of Business, 59(4), S251–S278. Tversky, A., & Kahneman, D. (1990). Rational choice and the framing of decisions. In: K.-S. Cook & M. Levi (Eds), The limits of rationality (pp. 60–89). Chicago: University of Chicago Press. Tversky, A., & Kahneman, D. (1991). Loss aversion and riskless choice: A reference dependent model. Quarterly Journal of Economics, 106, 1039–1061.
APPENDIX A In the simple case M ¼ 2, by solving the optimization problem (14) one can easily prove that demand for set 2 increases over time, i.e. qx(2)/qt>0 , if and only if Zsð2Þ ðtÞ4Zsð1Þ ðtÞ; where Zsð2Þ ðtÞ and Zsð1Þ ðtÞ are respectively the elasticity of s(2) and s(1)with respect to time, i.e. Zsð2Þ ðtÞ :¼ @sð2Þ =@t:t=sð2Þ and Zsð1Þ ðtÞ :¼ @sð1Þ =@t:t=sð1Þ : Ignoring the trivial case when both xð2Þ ox¯ ð2Þ and xð1Þ ox¯ ð1Þ hold, and the case of non-symmetric spillovers, i.e.j(1,2)6¼j(2,1) or ð1Þ absence of spillovers, i.e. j(1,2)6¼j(2,1) ¼ 0, if we assume that sð2Þ 0 ¼ s0 ; the previous condition is attained with certainty when set 2 is more creative than set 1, notably, when g(2)Zg(1) and b(2)Zb(1) and at least one of these inequalities is strong. In Figs. A1, A2a and A2b, x*(1) and x*(2) have been plotted under the assumption of fixed prices and fixed income. In Fig. A1 consumption set 1 at time t ¼ 0 is consumed above the minimum exposure threshold x¯ ð1Þ while we assume that xnt ð2Þ ox¯ ð2Þ ; for any tZ0, and f(2,1) 0. In Figs. A2a and A2b consumption set 1 is consumed above the threshold x¯ ð1Þ starting from t ¼ 0. In Fig. A2a we hypothesise demand function x*(2) never overcomes the minimum exposure threshold x¯ ð2Þ ; but there is positive spillover, in particular f(2,1) ¼ 0.5; in Fig. A2b, on the contrary, we suppose that set 2 is consumed above the threshold x¯ ð2Þ starting from t ¼ 0. In Fig. A3a and A3b set 1 is already consumed above the threshold x¯ ð1Þ at time t ¼ 0, whereas set 2 reaches the minimum exposure threshold setting x¯ ð2Þ ¼ 30 at time t ¼ 17.025. Before that time, of consumption
Idiosyncratic Learning, Creative Consumption and Well-Being
Fig. A1.
67
Demand functions where set 1 and set 2 are always consumed above and below the minimum exposure threshold respectively.
set 2’s hedonic exposure is realized by spillover (in particular we have f(2,1) ¼ 0.7); at time t ¼ 17:025 when the set 2 reaches quantity x¯ ð2Þ ¼ 30; the hedonic leaning function has a consequent jump and the accumulation of consumption human capital is complete. Fig. A3a shows the shape of the hedonic learning functions of both sets and Fig. A3b shows the corresponding demand functions. We plotted all previous figures setting a ¼ 6, Yc ¼ 100, p(1) ¼ 1.6, (2) p ¼ 1.9, s0ð1Þ ¼ s0ð2Þ ¼ 7; gð1Þ ¼ 3; gð2Þ ¼ 3:5; bð1Þ ¼ 8:5; bð2Þ ¼ 7:3:
APPENDIX B These results can be generalized to the case of more than two-set. In the latter case, in addition to subjective and objective creativity, total life wellbeing depends also on the number of goods available to generate distinct sets, in that the latter affects the opportunity to substitute more mature creative sets for less mature ones over an individual life. Notably, more variety implies more well-being as long as the available goods that form part of the sets differ sufficiently in terms of embodied characteristics and objective creative content (Figs. A4 and A5).
68
MARINA DI GIACINTO AND FRANCESCO FERRANTE
Fig. A2. (a) Demand functions where set 1 and set 2 are consumed from the beginning above the minimum exposure threshold. (b) Demand functions where both sets are consumed from the beginning above the minimum exposure threshold.
We plotted Figs. A5(a) and A6(a) setting a ¼ 6, Yc ¼ 100, p(1) ¼ 1.6, g(1) ¼ 3, g(2) ¼ 3.5, b(1) ¼ 8.5, b(2) ¼ 7.3, p ¼ 1.9, s0ð1Þ ¼ sð2Þ 0 ¼ 7; (1) (2) E ¼ E ¼ 0. Conversely, we plotted Figs. A5(b) and A6(b) setting ð2Þ (1) a ¼ 6, Yc ¼ 100, p(1) ¼ 1.6, p(2) ¼ 1.5, sð1Þ ¼ 3, g(2) ¼ 6.25, 0 ¼ s0 ¼ 7; g (1) (2) (1) (2) b ¼ 9, b ¼ 2, E ¼ E ¼ 0. (2)
Idiosyncratic Learning, Creative Consumption and Well-Being
69
Fig. A3. (a) Hedonic learning functions where set 1 and set 2 are consumed at the initial time above the minimum exposure threshold. (b) Demand functions in the same setup of (a).
APPENDIX C As we have observed, one may register both income and substitution effects due to an income increase (Due to the irreversible nature of learning, the
70
MARINA DI GIACINTO AND FRANCESCO FERRANTE W 12000 10000 8000 6000 4000 2000
(a)
20
40
60
20
40
60
80
t
100
W 12000 10000 8000 6000 4000 2000 t (b)
80
100
Fig. A4. (a) Well-being when the rates at which the consumption human capital of both sets have similar time valuation. (b) Well-being when the rates at which the consumption human capital of both sets have quite different time valuation.
non-standard effects of income variations are not symmetric i.e. a reduction of income does not yield substitution effects and reduces demand for all goods.) with negative net effect on demand for m. This scenario corresponds to the case discussed previously, of jumps in demand. In our 2-set setting, Figs. A6a, A6b, A7a and A7b depict a scenario in which there is a delay substitution effect in the case of income is Yc ¼ 100 initially and at t ¼ 5
Idiosyncratic Learning, Creative Consumption and Well-Being
Fig. A5.
71
(a) Hedonic learning function in the same setup of Fig. A4(a). (b) Hedonic learning function in the same setup of Fig. A4(b).
it jumps to Yc ¼ 130 (More precisely, consumption set 2’s hedonic exposure is obtained by spillover (imposing j(2,1) ¼ 0.25) from t ¼ 0 to t ¼ 5 to, while set 1 is consumed above the minimum exposure threshold starting ð2Þ from t ¼ 0 (for Yc ¼ 100, xð1Þ ¯ ð1Þ ¼ 19 and 0 ¼ x0 ¼ 19:2308 and we set x ð2Þ x¯ ¼ 20). Due to income change, at t ¼ 5 set 2’s hedonic learning is activated directly because its demand overcomes the minimum exposure
72
Fig A6.
MARINA DI GIACINTO AND FRANCESCO FERRANTE
(a) Demand function of set 1 when income is constant in time. (b) Demand function of set 1 when income increases at time t ¼ 5.
threshold (for Yc ¼ 130 we get x5ð2Þ ¼ 26:427; while for Yc ¼ 100 we would have xð2Þ 5 ¼ 16:9981).). We plotted the figures in this appendix setting a ¼ 6, p(1) ¼ p(2) ¼ 2.6, ð2Þ (1) sð1Þ ¼ 3.1, g(2) ¼ 3.3, b(1) ¼ 7, b(2) ¼ 8.5. 0 ¼ s0 ¼ 30; g
Idiosyncratic Learning, Creative Consumption and Well-Being
Fig. A7.
73
(a) Demand function of set 2 when income is constant in time. (b) Demand function of set 2 when income increases at time t ¼ 5.
This page intentionally left blank
PART B: UNCERTAINTY, NOVELTY, AND CHOICES
This page intentionally left blank
A SHACKLEAN APPROACH TO THE DEMAND FOR MOVIES John Sedgwick Film revenues are characterized by extreme variance, with a small number of very high-grossing movies generating a very long tail. The rates of return are similarly distributed: a handful of films generate extraordinary returns for their producer/distributors. However, there is no connection between the films that generate the highest revenues and those that earn the highest rates of return. Furthermore, film revenues are only imperfectly related to film costs: while an upward trend can be observed between the two variables, the magnitude of the variance rules out statistical significance and results in a random distribution of rates of return within any decile cost category.1 Such is the business environment of film producers, and it raises the question why this environment is so uncertain when film producers deliberately obtain and coordinate production inputs whose market value reflects idiosyncratic qualities that are high in demand. In answering this question it is interesting to reflect that a riskless environment for the producer/distributor would require (a) a direct, positive correspondence between the qualities of the set of artistic and technical inputs and the quality of the film produced by them and (b) the absence of any distinction between objective and subjective assessments of quality, so that audiences will always prefer films made from higher quality inputs; in other words, film production and film reception would be joined in a perfect coordination and matching of the producers’ assessments of quality The Evolution of Consumption: Theories and Practices Advances in Austrian Economics, Volume 10, 77–91 Copyright r 2007 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 1529-2134/doi:10.1016/S1529-2134(07)10004-1
77
78
JOHN SEDGWICK
with those of the consumers. The variance in the relationship between costs and revenues makes it clear that consumers do not behave according to those criteria, the reason being that they need to discover what it is that they like (De Vany & Walls, 1996, p. 1493). The problem facing producers is that of ‘reading’ and attempting to shape consumers’ tastes in the context of consumers not knowing fully what they want. The scope for wrong readings and disappointing discoveries is considerable. Hence, both producers and consumers are risk takers. The purpose of this paper is to link known empirical evidence of film popularity with a theoretical framework in which individual consumers use their imaginations to deal with the uncertainty associated with choosing the films they go to see theatrically. It is argued that whenever a sufficient number of consumers in general experience greater than expected cinematic utility from a particular movie, a positive information cascade causes the film to become a hit, joining, in the process, a small number of other films whose revenues form the thick long right tail of the revenue distribution. The mechanics of these information flows are well documented theoretically and empirically and are not the subject of this paper.2 Rather, this work proposes that consumers engage in a two-part decision-making process: the first stage involves the use of some form of personal heuristic to reject those films released onto the market that, for whatever reason, signal meagre levels of cinematic pleasure for the filmgoer. In the second stage, the consumer has arrived at a final choice set, consisting of a small number of films, which the consumer evaluates in a more considered fashion. In theorizing this final decision-making process, a Shacklean approach, as developed by Peter Earl (1993, 1995), is adopted, in which consumers conceive, only incompletely, a cinematic utility domain in which the expected utility promised by a film is mapped against the uncertainty of that utility being realized. While it is argued that filmgoers make errors when choosing between films, in that expectations and realization of utility frequently diverge, implying that they regularly mispredict utility, it is also argued that these errors are not systematic and indeed are randomly distributed. Further, the work does not call upon recent developments in consumer theory, closely associated with the literature on well-being, in which psychological experiments show that consumers in particular situations systematically make decisions that militate against their own best interest.3 Rather, the argument developed here is framed within the theoretical context of bounded rationality in which consumers (a) need to discover what it is they like, (b) cannot discover this before having had the experience, and (c) know from
A Shacklean Approach to the Demand for Movies
79
experience that previous cinematic encounters with markers such as actors or directors, and/or sequels, and/or genres are less than perfect guides to future cinematic utility.
FILM AS A COMMODITY-TYPE Film has a number of characteristics that define it as a commodity-type and hence distinguish it from other commodity-types.4 Not only is film nondiminishable in consumption (because it is consumed in the mind of the viewer), but also the images that make up a film are infinitely reproducible and, in the era of digital technology, reproducible at near zero marginal cost. Furthermore, each film is to some extent novel in that its constitutive images are each unique and so is the ordering of the images into the sequenced continuity that makes the film meaningful to (but not, ipso facto, liked by) the consumer.5 Hence, prior to the consumption of a film, consumers cannot have a complete idea of the visual and aural cinematic utility that they are going to experience, nor of their reaction to that experience. Films are thus experience goods: an experiential divide exists between the two mental states of awareness, namely expectation and realization, both of which entail a learning process based on previous experience (Nelson, 1974, p. 745). A final element in the ontology of film as a commodity is the rapidly diminishing utility of film in consumption. That is, once consumed in theatrical release, films are rarely revisited theatrically by consumers, who commonly prefer the anticipation of new cinematic pleasure to the repeat viewing of old pleasures. In forming their expectations, consumers interpret information signals made available by a film’s producers/distributors. These signals are likely to consist of markers such as actor or (less frequent) director as star, and genre-type, each of which, and in combination, becomes a factor within a consumer’s mental framework of evaluation, which in turn is a substructure within that person’s world-and-life-view (weltanschaunng) (Albert, 2005). It can be conjectured that consumers are likely to be attracted by signals that are consonant with the experiences that have formed their current tastes, and deterred or repelled by signals that are dissonant (Gilad, Kaish, & Loab, 1987). However, a consumer’s worldview is not an immutable monolith; although certain core components may change very little in a lifetime not beset with traumatic disruption, it is fundamentally a dynamic system that can be affected, exogenously, by political and economic circumstances (among which, employment, taxes, prices and incomes have immediate
80
JOHN SEDGWICK
Frequency
relevance to levels and patterns of consumption), technological innovation, the availability of leisure, and changing fashions in aesthetics and ideas, and, endogenously, by the ability of the consumer to learn through the consumption experience and be changed by it, perhaps imperceptibly in the short run, but noticeably over time. In this way, what may be dissonant information for a consumer today may prove to be more compatible with that person’s tastes in, say, 10 years time. Taste formation and hence expectations, thus involve both exogenous and endogenous aspects of change (Earl & Potts, 2004). The idea that consumers are subject to uncertainty in film consumption is captured by the curve in Fig. 1, which depicts a hypothetical ex post frequency distribution of the difference between a consumer’s expectation and realization of cinematic utility derived from each film seen over a lifetime. Thus, every new act of film consumption adds to the distribution, so that if the experience is positive for the consumer, it adds to the right-hand-side of the distribution, and vice versa. This is not to say that the state of expectation is a steady state – far from it. As argued above, both endogenous and exogenous factors constantly influence it, causing it to be in a permanent state of flux. Nevertheless, from the fact that a consumer, at any moment of time, is able to form an expectation, however loosely defined, of the anticipated cinematic utility promised by a film, it follows that each instance of film consumption must result in one of three possibilities, an experience of: (1) diminished utility, or (2) fulfilment, or (3) enhanced utility.
Region of Diminished Utility
Fig. 1.
F
Region of Enhanced Utility
Frequency Distribution of the Difference between Expectation and Realization of Cinematic Utility.
81
A Shacklean Approach to the Demand for Movies
Diminished utility can run the gamut from mild disappointment or sense of loss to the depths of dispiritedness and/or obfuscation; conversely, enhanced utility can range from a slight gain in pleasure to the heights of exhilaration and/or revelation. The curve in Fig. 1 represents the (hypothetical, idealized) diachronic lifetime (say, n-year) totality of one consumer’s difference-between-expectationand-realization experiences.6 This would be the result of the data obtained from a hypothetical longitudinal study of the consumer, from year 1 to year n. Each year’s data, taken separately would give a ‘snapshot’ of the consumer’s expectation experiences, but when added year-on-year, the data would provide a record of the diachronic evolution of the consumer’s difference-between-expectation-and-realization experiences as revealed by the changes in kurtosis through the years. A real-life longitudinal investigation would, of course, study a population of consumers, and thus be able to produce comparative synchronic ‘snapshots’ and diachronic evolutions, as proposed in Fig. 2. It would be possible to see how the kurtoses of the distributions differed at various junctures in the lives of the consumers, and whether, as might be supposed, young teenagers and middle-aged and elderly filmgoers have more conservative tastes and are less prepared to take risks when choosing between films, causing the kurtoses of their distributions to be higher than those of their young adulthood, when they were more prepared to experiment.
Frequency
Conservative stage
Most adventurous stage
Region of Diminished Utility
F
Region of Enhanced Utility
Fig. 2. Frequency Distributions of the Difference between Expectation and Realization of Cinematic Utility at Various Stages in a Consumer’s Life.
82
JOHN SEDGWICK
A THEORY OF FILM CHOICE The distribution of the differences (discrepancies) between a filmgoer’s expectations and experiences, which is the subject of Fig. 1, is meant as an illustrative conception of the filmgoer’s ex post viewing history. It tells us that a consumer’s personal history of filmgoing is an imperfect guide to future film quality, and hence, ex ante, filmgoers cannot but endure uncertainty. As Hoskins, McFadyen, and Finn (1997, p. 56) have argued: A feature film is a product that consumers must pay for before they know how much enjoyment they will receive. Attending large numbers of films provides little guidance in choosing a new movie y In other words, search activity and experience, which are valuable to consumers for many other products, are of little guidance to consumers in choosing which movie to attend.
Key to understanding the uncertainty faced by filmgoers is their inability to conceive ex ante the full range and intensity of sensory experiences entailed in the act of film consumption. Their conception is necessarily hazy – it is incomplete. Consequently, it is not possible that consumers can ascribe probability values to the range of possible outcomes. For this reason, standard expected utility theory is not an appropriate theoretical framework for understanding consumer decision-making. As George Shackle (1948, p. 38) has argued: Hopes which are mutually exclusive are not additive; fears which are mutually exclusive are not additive. In each case the greatest prevails, and alone determines the power of the attractive or of the deterrent component of the venture’s ‘‘dual personality’’. In this last sentence, the word ‘‘greatest’’ is insufficiently precise y What we mean is the most powerful element among them.
Filmgoers cannot ‘assess risk’ in the technical sense of the algorithms used by insurance company statisticians, epidemiologists, etc. In proposing an alternative theoretical framework for understanding filmgoers’ choices, this paper makes use of Peter Earl’s (1995, pp. 112–118) adaptation of Shackle’s ‘theory of surprise’, in which the state of being surprised is an everyday occurrence entailed by our not having a complete conception of events – of their causes, their nature, and the consequences of action taken in relation to them. Let us assume that potential audiences develop heuristic practices, based on their preferences for particular markers, such as actor and, or, director as star, genre, production value, storyline, and emotional appeal, for filtering films from their choice set, in order to arrive at a decision set. Fig. 3 depicts a continuum in which the scale of expected cinematic utility promised by films
83
A Shacklean Approach to the Demand for Movies
REJECT SET F3 F2 F1
DECISION SET F4
Low levels of expected cinematic utility
Fig. 3.
F5
F6
High levels of expected cinematic utility
The Initial Decision-Making Process.
is ranged from left to right in ascending order, so that, for the i-th potential consumer, film F6 promises higher levels of cinematic utility than film F5, which in turn offers more than film F4, and so on. However, while films F5 and F6 form the decision set, films F1–F4 are rejected from further consideration. Empirical support for this type of decision-making behaviour can be derived from the extremely unequal distribution of film revenues that characterizes the film industry, from which it is evident that during the course of a year large numbers of filmgoers go to see a relatively small number of films. Further, the rapid life and death process of films on theatrical release means that during any single week, filmgoers are faced with a handful of films to choose between, of which they are likely to have heard and/or read about only a few. With the decision set consisting of, say, two films F5 and F6, the consumers must go on to make a more informed reflective assessment of the merits of F5 and F6. Using the Shacklean framework, let us suppose that consumers are able to form expectations and, given their personal histories, imagine their reaction to the extent to which those expectations may be met, on a scale from ‘perfectly possible [no surprise]’, to ‘astonishment’. In other words, the Shacklean consumer, reviewing the possible outcomes of expectation mentioned above – diminished utility, fulfilment, or enhanced utility – asks ‘how surprised would I be?’ Thus a filmgoer could be astonished by how dispiriting or, conversely, how revelatory the experience of watching a particular film has been. Astonishment, however, is a comparatively rare phenomenon. More likely, is the state of ‘not being much surprised’ by the cinematic pleasures derived from films. Indeed, it is assumed that consumers are able to map expectations in such a way as to allow them to form a judgement about what is the most likely best-bet outcome, about which, when it happens, they are not surprised at all. Set within the context of the microtheory built around Fig. 1, all of the films consumed over a filmgoer’s adult lifetime are decision-set films that
84
JOHN SEDGWICK
were selected ahead of, or at the expense of, the other films in that person’s choice sets. Furthermore, the spread of the distribution suggests that a relatively small number of films had surprised the consumer greatly: they had either seriously failed to meet, or greatly exceeded, what was expected of them, in that they had resulted in large negative or large positive differences between expectation and realization. Such films may have astonished filmgoers, either in getting their choice so wrong, or in having their senses aroused to an extraordinary, but unexpected degree. Knowing this to have happened to them, it may be supposed that a filmgoer is able to imagine a range of possibilities for loss and gain set against degrees of being surprised to some extent. In Fig. 4 these ideas are examined. Imagine that filmgoers are able to make sense of uncertainty through the anticipation of surprise associated with a given action and that they conceive the degree of surprise along a continuum ranging from perfectly unsurprising to astonishing, depicted in Fig. 4 by the vertical axis. The degree of surprise represents the state of uncertainty associated with film consumption – the impressionistic ‘strength’ of the filmgoer’s subjective awareness of the difference between expectation and realization. Now suppose that two films, A and B, make up an individual filmgoer’s decision set, and that, in considering them, the filmgoer is unable to achieve a clinching, exclusive, ‘must-see’ level of expectation for one, but instead arrives at the same high, ‘well-worth-taking-a-chance-on’ level of best-bet expectation E for both, so that the filmgoer cannot distinguish between them in terms of the degree of cinematic utility alone that
Degree of Potential Surprise
Film B
Film A
d2 d1 Prospective Enhanced Utility E e1 e2 Prospective Diminished Utility Decision-set part of the Continuum of Cinematic Utility
Fig. 4. A Consumer’s Ex Ante Choice Situation: Two Films that have Identical ‘Well-Worth-Taking-a-Chance-on’ Expectation Levels, but have Different Surprise Profiles.
85
A Shacklean Approach to the Demand for Movies
Degree of Potential Surprise
each film promises.7 With disposable time and money, the filmgoer could of course avoid the fork and see both films, but this discussion is concerned with ways of making choices, not evading them, so in this situation the decision will rest on the ‘surprise’ factor. However, it is apparent that the level of potential surprise with respect to E, as depicted by the two surprise curves, is highly sensitive to what is chosen, with Film A being much more difficult to call in relation to the consumer who understands that there would be only marginally more surprise in experiencing considerably diminished cinematic utility, say at point d2, or, its opposite state, greatly enhanced cinematic utility, say e2, than in experiencing only small negative, d1, or positive, e1, departures from expectation E. Under these circumstances, the decision about what film to choose would depend upon the filmgoer’s attitude to loss and gain of cinematic utility. Thus although indifferent as between the two films at point E, the filmgoer who values the marginal utility of gain above the marginal utility of loss will choose Film A, and Film B if the reverse is the case. Fig. 5 presents a more interesting scenario. Film A and Film B are again characterized by surprise curves similar to those found in Fig. 4. However, on this occasion they generate different best-bet expectations, with Film B promising a higher level of cinematic utility than Film A, i.e., EB4EA. In this situation, the decision will depend on the filmgoer’s degree of tolerance of surprise. If at the margin, gain were valued more highly than loss, the decision would depend upon just how much uncertainty the filmgoer was willing to endure. A gain-loving filmgoer in search of revelatory experiences who was prepared to tolerate surprise beyond the level s1 would gamble on
Film B
Film A
s1 Prospective Diminished Utility
EA
EB
Prospective Enhanced Utility
Decision-set part of the Continuum of Cinematic Utility
Fig. 5. A Consumer’s Ex Ante Choice Situation: Two Films have Different ‘WellWorth-Taking-a-Chance-on’ Expectation Levels and Different Surprise Profiles.
86
JOHN SEDGWICK
Film A, even if the potential for diminished utility were considerable. On the other hand, a gain-loving filmgoer less tolerant of surprise would choose Film B instead of Film A, as would the loss-hating filmgoer. Clearly, at levels of surprise below s1, Film B dominates.8 These choice possibilities are set out in Table 1, leading, it can be conjectured, to the identification of three types of filmgoers – when defined by attitude towards diminished cinematic utility and uncertainty. First, there is the conservative filmgoer, who is attracted to standard quality markers such as stars, production values, and generic conventions, which are likely to give rise to films with a profile similar to that of Film B: a person who is not very tolerant of films for which the potential for diminished cinematic utility is considerable. Second, there is the filmgoer who, unlike the first, is willing to experiment and does not hold the standard quality markers in such high regard, but, like the first filmgoer, is not very tolerant of diminished cinematic utility. Again, Film B is likely to be chosen. Third, there is the filmgoer who, like the second, is willing to experiment and does not hold the standard quality markers in such high regard, but, unlike the second filmgoer, discounts the potential for diminished cinematic utility in favour of the chance of experiencing an elevated level of enhanced cinematic utility. In this case, it is likely that Film A will be chosen. This analysis casts some light on the production strategies of the major Hollywood studios, in that they need to produce films that (a) generate very high levels of expectation so that they get into consumers’ decision sets en Table 1.
The Filmgoer’s Uncertainty Matrix. Filmgoers prepared to tolerate only low levels of potential surprise – an unwillingness to experiment
Filmgoers prepared to tolerate only low levels of diminished cinematic utility Filmgoers who give more weight to the promise of elevated levels of enhanced cinematic utility than to the possibility of diminished cinematic utility
Filmgoers prepared to tolerate high levels of potential surprise – a willingness to experiment
Film B
Film B
Film A
A Shacklean Approach to the Demand for Movies
87
masse and (b) reduce to a minimum the degree to which consumers believe that they may experience diminished cinematic utility. Thus, film studios ideally aim to make films that generate asymmetrical surprise curves for the filmgoer: in the region of diminished utility, a curve with a steep arm close to the surprise-axis, turning about the E point into a flatter arm in the region of enhanced utility, offering low chances of diminished utility with the distinct possibility that the film will prove to be a momentous occasion for its audiences. Clearly, advertising, star and production values play an important role here in shaping these expectations. However, filmgoers, with their wealth of experience, are much more likely to form steep ‘U’-shaped curves when considering the prospect of going to see a screening of a standard Hollywood production. It is only the hit films of the season that signal, through ‘opening’ filmgoers’ large-scale positive word of mouth, asymmetrical curves of the sort described above for subsequent cohorts of filmgoers (Moul & Shugan, 2005). Occasionally, a film has a Top 20 run that surprises everybody. An example of this, in 1998, was the Italian (language) film Life is Beautiful (La Vita e` Bella). Made with a budget of US$6.5 million, the film started its US distribution at six cinemas in October of that year. By the end of its release it had grossed over US$57 million. Entering the Top 20 chart at the beginning of November, in 19th position, the film remained in the lower half of the charts for the next 20 weeks, taking up a Top 10 berth for two weeks in April 1999 before finally leaving the charts in June of the same year. For five consecutive weeks, during the weeks of late March and April, the film was screened at over 1,000 cinemas. Films such as Life is Beautiful are known to the trade as ‘sleepers’, films that build their audiences over time, significantly going beyond the original business concept conceived for it by its producer/ distributor. In 1998, the median length of run of the 185 films that on release achieved at least one week’s berth in the weekly Top 20 charts was just four weeks! Contrast this performance with two contemporary Hollywood big budget productions, Patch Adams and Saving Private Ryan, which grossed US$135 million and US$216 million, opening on 2,712 and 2,463 screens, respectively. These two films on initial release would have been in many more filmgoers’ decision sets than Life is Beautiful, which would not have been on most filmgoers’ mental radar – it would have been languishing near F1 in Fig. 3. It can be conjectured that for those that did go to see the film during the early period of its release, the film would have generated a surprise curve similar to that of Film A in Fig. 5, whereas the two Hollywood films would more likely have caused potential filmgoers to conceive of them in terms of
88
JOHN SEDGWICK
Saving Private Ryan
Patch Adams
Life Is Beautiful
3,500 3,000
Screens
2,500 2,000 1,500 1,000 500
43
40
37
34
31
28
25
22
19
16
13
10
7
4
1
0 Weeks in Weekly Top 100
Fig. 6. The Diffusion Patterns of Life is Beautiful, Patch Adams, and Saving Private Ryan. Note: Saving Private Ryan was all but removed from circulation during the 11 weeks preceding the Oscar awards in February 1999, at which the film won Best Film category. Source: Variety.
Film B-type curves. The diffusion pattern of Life is Beautiful, contrasting strongly with the two A-feature Hollywood productions, is captured in Fig. 6. Clearly, in the minds of mainstream filmgoers, the film worked its way from barely-on-their-radar to decision-set status – from F1 to F5 or F6 in terms of the nomenclature of Fig. 3.
CONCLUSIONS This paper proposes a theory of film choice based analytically upon the distinction between the ex ante set of expectations that consumers bring to the decision-making process and the ex post experience of having seen the chosen film. The latter leads us to suppose that, over a lifetime, filmgoers construct a mental ledger of their experiences, from which they form an empirical distribution of the extent to which expectations have been fulfilled, or otherwise. The accumulation of experience thus serves as an ongoing basis for decision-making. Of course, a whole host of exogenous factors, including the marketing efforts of studios to influence the filmgoer, also
A Shacklean Approach to the Demand for Movies
89
affects the consumer’s choice mechanism, a further examination of which leads to the argument that the inescapable conceptual incompleteness with which consumers approach new films makes inappropriate the standard probability-based approach to consumer theory. The paper proposes that Shackle’s theory of surprise is a more fitting framework for examining consumer choice in that filmgoers bring an experiential history to the choice situation from which they are able to form decisive expectations by imagining outcomes.
NOTES 1. See De Vany and Walls (1996), Sedgwick (2002), Sedgwick and Pokorny (2005), and Walls (2005) for analysis of the peculiar nature of film revenues at different historical junctures. 2. For this see Bikhchandani, Hirschleifer, and Welch (1992), De Vany and Walls (1996, 2004). 3. See Frey and Stutzer (2002) and Kahneman, Wakker, and Sarin (1997). 4. For a full discussion of the ontology of film as a commodity see Sedgwick (2000, pp. 7–16). 5. See Bianchi (1998, 2002) for a discussion of the importance of novelty in consumption. 6. This is a simple, idealized presentation for the purposes of discussion. An empirical long-term longitudinal investigation of a consumer’s lifetime of film viewing expectations would no doubt reveal a curve with skew, multimodality, and other complications. 7. Surprise occurs as a response to a departure from what was expected. It can vary in degree or intensity from the barest emotional awareness of a discrepancy to the shock of astonishment. If, in doing or experiencing something, there is no departure from expectation (i.e., the expectation is fulfilled) there is no surprise. It follows, therefore, that when, in choosing a film, a filmgoer has arrived at a bestbet expectation, E, of cinematic utility, the graph of the filmgoer’s Potential Surprise against Cinematic Utility will show a curve with a turning point touching the Utility axis at the point (E, 0). 8. For the sake of exposition, the surprise curves in Figs. 4 and 5 have been drawn as symmetrical about their respective best-bet E points. However, as the curves will be unique for each individual, ex ante the films might promise asymmetrical distributions of potential utility, depending on the tastes and perceptions of the filmgoer.
ACKNOWLEDGMENTS I would like to thank Bernard Hrusa Marlow for his painstaking assistance in clarifying my ideas and Marina Bianchi, Gherardo Girardi and
90
JOHN SEDGWICK
Guglielmo Volpe for their thoughtful reviews of this work, as well as the participants of the ‘Evolution of Consumption Workshop’ organized by the University of Cassino and held in Gaeta, Italy in March 2005 and the ‘7th Annual Business and Economics and Scholars Workshop Summit in Motion Picture Industry Studies’, held at the Carl De Santis Business and Economics Center for the Study and Development of the Motion Picture Industry, at Florida Atlantic University, in November 2005.
BIBLIOGRAPHY Albert, S. (2005). Movie stars and the distribution of financially successful films in the motion picture industry. In: J. Sedgwick & M. Pokorny (Eds), An economic history of film. London: Routledge. Bianchi, M. (1998). The taste for novelty and novel tastes. In: M. Bianchi (Ed.), The active consumer: Novelty and surprise in consumer choice (pp. 64–86). London: BFI. Bianchi, M. (2002). Novelty, preferences and fashion: When goods are unsettling. Journal of Economic Behaviour and Organization, 47, 1–18. Bikhchandani, S., Hirschleifer, D., & Welch, I. (1992). A theory of fads, fashion, custom, and cultural change as informational cascades. Journal of Political Economy, 100, 992–1026. De Vany, A., & Walls, W. D. (1996). Bose-Einstein dynamics and adaptive contracting in the motion picture industry. Economic Journal, 106, 1493–1514. De Vany, A., & Walls, W. D. (2004). Big budgets, big openings and legs: Analysis of the blockbuster strategy. In: A. De Vany (Ed.), Hollywood economics: How extreme uncertainty shapes the film industry. London: Routledge. Earl, P. (1993). The economics of G.L.S. Shackle in retrospect and prospect. Review of Political Economy, 5, 127–137. Earl, P. E. (1995). Microeconomics for business and marketing. Aldershot: Edward Elgar. Earl, P. E., & Potts, J. (2004). The market for preferences. Cambridge Journal of Economics, 28, 619–633. Frey, B., & Stutzer, A. (2002). Happiness and economics: How the economy and institutions affect well-being. Pinceton and Oxford: Princeton University Press. Gilad, B., Kaish, S., & Loab, P. (1987). Cognitive dissonance and utility maximization: A general framework. Journal of Economic Behaviour and Organization, 8, 61–73. Hoskins, C., McFadyen, S., & Finn, A. (1997). Global television and film. Oxford: Oxford University Press. Kahneman, D., Wakker, P., & Rakesh, S. (1997). Back to Bentham? Explorations of experienced utility. Quarterly Journal of Economics, 112, 375–405. Moul, C. C., & Shugan, S. M. (2005). Theatrical release and the launching of motion pictures. In: C. C. Moul (Ed.), A concise handbook of movie industry economics. Cambridge: Cambridge University Press. Nelson, P. (1974). Advertising as information. Journal of Political Economy, 82, 729–745. Sedgwick, J. (2000). Popular filmgoing in 1930s Britain: A choice of pleasures. Exeter: Exeter University Press.
A Shacklean Approach to the Demand for Movies
91
Sedgwick, J. (2002). Product differentiation at the movies: Hollywood, 1946–65. Journal of Economic History, 62, 676–704. Sedgwick, J., & Pokorny, M. (2005). The film business in the U.S. and Britain during the 1930s. Economic History Review, 58, 79–112. Shackle, G. (1948). Expectations in economics. Cambridge: Cambridge University Press. Walls, W. D. (2005). Modelling movie success when ‘nobody knows anything’: Conditional stable distribution analysis of film returns. Journal of Cultural Economics, 29, 177–190.
This page intentionally left blank
THE EVOLUTION OF ENTERTAINMENT CONSUMPTION AND THE EMERGENCE OF CINEMA, 1890–1940 Gerben Bakker 1. INTRODUCTION At the end of the nineteenth century, in the era of the second industrial revolution, falling working hours, rising disposable income, increasing urbanisation, rapidly expanding transport networks and strong population growth resulted in a sharp rise in the demand for entertainment. Initially, the expenditure was spread across different categories, such as live entertainment, sports, music, bowling alleys or skating rinks. One of these categories was cinematographic entertainment, a new service, based on a new technology. Initially it seemed not more than a fad, a novelty shown at fairs, but it quickly emerged as the dominant form of popular entertainment. This paper argues that the take-off of cinema was largely demand-driven, and that, in an evolutionary process, consumers allocated more and more expenditure to cinema. It will analyse how consumer habits and practices evolved with the new cinema technology and led to the formation of a new product/service. Two questions are addressed: why cinema technology was introduced in the mid-1890s rather than earlier or later; and why cinema-going became The Evolution of Consumption: Theories and Practices Advances in Austrian Economics, Volume 10, 93–137 Copyright r 2007 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 1529-2134/doi:10.1016/S1529-2134(07)10005-3
93
94
GERBEN BAKKER
popular only with a lag – a decade after the technology was available. Both issues can potentially be affected by changes in supply or changes in demand. Section 2 sets both alternatives against a more detailed history of innovation and the emergence of cinema consumption, sharpening our sense of both the technology aspect and the lag between technical possibility and take-off. In Section 3 the available data sources relevant to understanding how the consumption of cinema grew are identified and analysed in depth and national differences decomposed in those due to technology and those due to taste. Section 4 further investigates the demand-led explanation of the emergence of cinema by locating it within the changing demand for recreational spending as a whole.
2. THE EVOLUTION OF FILM PRODUCTION 2.1. The Lag between Technology and Innovation As with many innovations, the idea of cinema preceded the invention itself. It is difficult to give an exact date to the emergence of the idea, or concept of cinema, but the first projection of moving images dates from the 1850s, and the first patents on the viewing and projection of motion pictures were filed in 1860. The more specific idea of applying all these ideas into one technology must have emerged at least some time before the mid-nineteenth century (Michaelis, 1958, pp. 734–751; 734–736). Many visual devices and gadgets preceded cinema, too many to list here in detail. A widespread and well-known one was the camera obscura, first constructed in 1645, which projected views in a dark room, for painters. Around the same time Anastasius Kircher built a special room to project images with mirrors, which looked somewhat like a cinema. A specialised building with many people using specialised equipment was necessary to project the images. About a decade later, in 1659, the Dutchman Christiaan Huygens invented the magic lantern, an easy, portable device, which could project images painted on a glass plate. Huygens’ interest was mainly scientific, but in the 1660s, the first showman, Thomas Walgensten, a Danish teacher and lens grinder living in Paris, travelled Europe giving exhibitions of the marvellous magic lantern. Not much later, a vibrant business of travelling showman, equipment manufacturers and slide painters emerged. At least from the 1740s onwards, magic lantern shows were also given regularly in the US (Musser, 1990, pp. 17–20).
95
The Evolution of Entertainment Consumption
In 1799, the Frenchman Etienne Gaspart Robert became well-known for his spectacular shows with magic lanterns in Paris, which he named the Fantasmagorie. Robert used several projectors, moved by operators to get larger and smaller images, smoke, sound effects and many other tricks and gadgets. The audience saw, for example, a ghost becoming larger and larger as if it was flying into the audience and then at the last moment disappear. In the early 1800s, Robert and his Fantasmagorie also travelled to Britain and the United States, where he asked a one dollar entry fee (Musser, 1990, pp. 24–25; Michaelis, 1958, pp. 736–737). Cinema as it was introduced in the late 1890s, was based on seven important technologies, ideas or concepts (Table 1). First, it was based upon photography, invented in the 1830s. It was also based upon two further innovations in photography. The separation of making photographs by first taking pictures on a negative, and only later making as many positives as one wants, was the second important innovation for cinema technology, as it enabled duplication and it made faster picture-taking possible. This innovation took place in the late 1880s, and became the industry standard quickly after the introduction of the Kodak pocket camera by George Eastman (Ko¨nig & Weber, 1990, pp. 527–530). The third innovation, the roll film made it possible to take many pictures – a hundred in the first Kodak camera – without having to change film. Experiments with roll film started in the 1850s, and it became the standard with the introduction of the Kodak camera (Ko¨nig & Weber, 1990, pp. 527–530). Fourth, celluloid was important. The first Kodak roll films used paper as a base, but since film cameras use large rolls, paper was not strong and reliable enough to serve as a base. Invented in 1868 and available in sheet form since 1888, celluloid could do the task, although for film-cameras The Technologies of Cinema, 1645–1888.
Table 1. Technology
When available In principle
Photography Positives and negatives Roll films Celluloid base High sensitivity emulsion Projection Dissection/persistence of vision
1850s 1868
1645 1826, 1872, 1874
Inventor
Alternatives
Innovation 1830s Late 1880s 1888 1888 Late 1880s 1851 1895
Kodak Kodak Goodwin/Kodak
Drawings/cartoons Positive–positive Cylinders with paper Paper base Low sensitivity emulsion with longer exposure Peep-hole machines Continuous photography (CCD-microchips)
96
GERBEN BAKKER
thicker strips of celluloid were used than for photo-cameras (Friedel, 1979, pp. 45–62; Michaelis, 1958). Fifth, a major obstacle for the invention of the motion-picture camera was the low sensitivity of the photographic emulsion, which made it impossible to take pictures at high speed, and thus to film motion. For the early portraits, people had to sit still for several seconds, and for motion pictures this simply could not be done. In the late 1880s, when new emulsions were tried, the sensitivity of film finally was so much improved that minimum length of exposure sufficiently shortened to make motion picture taking possible (Musser, 1990, pp. 45, 65). Sixth, the concept of projection was important for motion pictures, although in the original Edison invention, projection was lacking. In 1851, onwards, when the projection of photographic slides became possible, the magic lantern became wildly popular, and the industry started to grow quickly (Michaelis, 1958; Musser, 1990, pp. 30–36). A few specialised British and French slide suppliers dominated the trade. They collected photographs from all over the world in London or Paris, and distributed them quickly again to all corners of the globe. The largest firm was probably the French Levy and Company, which was acquired by the American firm of Benerman and Wilson in 1874. The photographic lantern slides enabled people to get used to sitting in a room and watching pictures of far away places, and for the first time to seeing pictures of news events that they had read about (Michaelis, 1958; Musser, 1990). Seventh, the idea of slicing a view with movements into small dissections, each of a fraction of a second, combined with the idea that when this would be shown the audience would see the movement because of the persistence of vision, was important to cinema. The notion of the persistence of vision is old, and was used in several of the visual gadgets of the nineteenth century, such as the Thaumatrope and the projection of a cartoon. The idea to dissect a view, however, was newer, and started with the photographs of Marey to capture the movement of horses in 1872, followed by the American Muybridge in the same year. The astronomer Jansen used the concept in 1874 to make observations of Venus. 2.1.1. The Innovation Process After the preconditions for motion pictures had been established, cinema technology itself was invented. Already in 1860/1861 patents were filed for viewing and projecting motion pictures, but not for the taking of pictures. The scientist Jean Marey completed the first working model of a film camera in 1888 in Paris. That year, Edison visited Marey and watched his films.
The Evolution of Entertainment Consumption
97
In 1891, Edison filed an American patent for a film camera, which had a different moving mechanism than the Marey camera. In 1890, the Englishman Friese Green presented a working camera to a group of enthusiasts. In 1893, the Frenchman Georges Demeney filed a patent for a camera. Finally, the Lumie`re brothers filed a patent for their type of camera and for projection in February 1895. In December of that year, they gave the first projection for a paying audience. They were followed in February 1896 by the Englishman Robert W. Paul. Paul also invented the ‘Maltese cross’, a device still used in cameras today, and instrumental in the smooth rolling of the film (Michaelis, 1958; Musser, 1990, pp. 65–67; Low & Manvell, 1948). Several characteristics stand out in the innovation process. First, it was an international process that took place in several countries, the inventors building and improving upon each others’ inventions. This fits with Mokyr’s notion that in the nineteenth century innovations increasingly came to depend on international communication between inventors (Mokyr, 1990, pp. 123–124). Second, it was what Mokyr calls a typical nineteenth century invention, in that it was a smart combination of many existing technologies. Many different innovations in the technologies, which it combined, had been necessary to make possible the innovation of cinema. Third, cinema was a major innovation because it was quickly and universally adopted throughout the western world, quicker than the steam engine, the railroad or the steamship. To sum up, the basic constituent technologies were all available in 1888, while the first working innovation was produced three years later, in 1891, and the ‘stable’ innovation seven years later, in 1895. This shows a time lag, albeit it a rather short one. The time lag is long enough, however, to allow us to retain the hypothesis that the invention of cinema was largely demandled, but it is so short as to leave a lot of doubt and calls for the other tests to show more conclusive outcomes, if the null hypothesis (cinema was a supply-led invention) is to be rejected.
2.2. The Lag between Innovation and Take-Off 2.2.1. The Take-Off of the Film Industry/Growth Phases For about the first 10 years of its existence, cinema in the United States and elsewhere was mainly a trick and a gadget. Before 1896 the coin-operated Kinematograph of Edison was present at many fairs and in many entertainment venues. Spectators had to throw a coin in the machine and peek
98
GERBEN BAKKER
through glasses to see the film. The first projections, from 1896 onwards, attracted large audiences. Lumie`re had a group of operators who travelled around the world with the cinematograph, and showed the pictures in theatres. After a few years, around 1900, films became a part of the programme in vaudeville and sometimes in theatre as well. Also, around 1900, travelling cinema emerged: cinemas, which travelled around with a tent of mobile theatre and set up shop for a short time in towns and villages. These differed from the Lumie`re operators and others in that they catered for the general, popular audiences, while the former were more upscale parts of theatre programmes, or a special programme for the bourgeoisie (Musser, 1990, pp. 140, 299, pp. 417–420). This era, which in the US lasted up to about 1905/1906, was a time in which cinema seemed just one of many new fashions, and it was not at all sure that it would persist. This changed between 1905 and 1907, when Nickelodeons, fixed cinemas with a few hundred seats, emerged and quickly spread all over the country.1 It can be said that from this time onwards cinema changed into an industry in its own right, which was distinct from other entertainments, since it had its own buildings and its own advertising. The emergence of fixed cinemas coincided with a huge growth phase in the business in general; film production increased greatly, and film distribution developed into a special activity, often managed by large film producers. However, until about 1914, besides the cinemas, films also continued to be combined with live entertainment in vaudeville and other theatres (Musser, 1990; Allen, 1980). We can thus place the take-off of the cinema industry between 1905 and 1907. In these years it developed its own retail outlets and did not depend exclusively on theatres and travelling showmen. From this time onwards the business also came to be seen as more than just a fad or fashion like skating rinks and bowling alleys. At the same time an increase in its growth pace started: it began to grow very fast, and slowly but gradually some people substituted cinema for small-time vaudeville and ‘popularpriced theatres’. Fig. 1 shows the total length of negatives released on the US, British and French film markets. The US time-series go back the farthest, giving an opportunity to analyse the early growth of the industry. Clearly, the initial growth between 1893 and 1898 was very strong, albeit from a very low initial base – the market increased with over three orders of magnitude. Between 1898 and 1906, far less growth took place, and in this period it may well have looked like the cinematograph would remain a niche product, a gimmick shown at fairs that used to be interspersed with live
99
The Evolution of Entertainment Consumption
Total released length (meters)
10,000,000
1,000,000 UK 100,000 I
10,000
FR
1,000
100 US 10 1890
1895
1900
1905
1910
1915
1920
Fig. 1. Total Released Film Negative Length, US, UK, France and Italy, in Meters, 1893–1922. Note: See Bakker (2005a, 2005b, Appendix I) for the method of estimation and for a discussion of the sources. Source: Bakker (2001b); American Film Institute Catalogue 1893–1910; Motion Picture World 1907–1920; Cine Journal 1908–1923. French data between 1901 and 1907 have been obtained by calculating a weighted growth index from the growth indices of Gaumont (1/3) and Pathe´ (2/3) of their released negative length (as reported in Meusy, 2002: p. 427). This growth index is then linked to the Cine Journal length-series and used to compute length from 1901 to 1907. The years 1908 to 1910, for which both datasets are available, suggest that the growth rates are quite comparable, although not exactly the same. Italian data from Redi (1995), as quoted in Meusy (2002, p. 420).
entertainment. From 1906, however, a new, sharp, sustained growth phase starts, with the market increased further again, by two orders of magnitude – and from a far higher base this time.2 During the interval in which time series overlap, the British and French negative length was growing at roughly the same rates as the US one, until 1914. That war year constitutes a great discontinuity, and from then on European growth rates are different and far lower than US ones. At the same time, the average film length increased considerably, from 80 feet in 1897 to 700 feet in 1910 to 3,000 feet in 1920. As a result, the total released length, which is the best indicator of production, increases more
100
GERBEN BAKKER
rapidly than the number released, in the US from 38,000 feet in 1897, to 2 million feet in 1910, to 20 million feet in 1920. 2.2.2. Emergence of Cinema Consumption Representative audience surveys on early motion picture audiences are lacking, and modern market research was not yet done by the emerging movie companies (Bakker, 2003). The only information available is from the press and trade press and from company sources. Before the era of fixed cinemas emerged, probably a dual audience existed. At the high end was the upper middle class, who saw the first shows of Lumie`re’s cinematograph probably in a legitimate theatre, as a special event, and later on between the live-acts in big-time vaudeville. At the other end, a more mixed social cross-section of local communities came to see the travelling cinema when showmen visited their town. This audience probably came from all layers of the population (Musser, 1990, pp. 140, 417–420). In the US, once the Nickelodeons had emerged between 1905 and 1907, their audience seems to have been mixed. Women and children probably constituted about half of the audiences and they might even have been the majority of visitors. Richard Abel relates, for example, that in New York, women often went with their children to the Nickelodeon after or during shopping, as these venues were handily located in the shopping districts (Abel, 1999, p. 48). A substantial difference between cinema and many other entertainments was that cinema was consumed by members of both sexes, while football, other sports, drinking and music hall were mostly an all-male event. When women were allowed in music halls, it was on the galleries, separated from the men. Compared to the previous entertainments, cinema was thus a whole new experience for consumers (Bakker, 2001a).3 Garth Jowett (1974) distinguishes three major audience groups: the middle classes that had never attended theatre or other amusements because of religious beliefs; the middle and upper working-class patrons of the live theatre, especially fans of popular melodramas; and the large urban working class who seldom went near theatrical entertainment. Some estimates put 78 per cent of the New York audience in the latter group (Jowett, 1974). Little is known about the age of the cinemagoers. The intuition is that they were mainly below the age of 30 or 40 (Abel, 1999, p. 48). Even so, little is known about the frequency of visits. People who happened to live or work near a Nickelodeon would probably visit it once a week, and other people less frequently. The audience is generally thought to be the less well-off classes, and immigrants who had difficulty with the English language and therefore were a natural market for motion pictures (Musser,
The Evolution of Entertainment Consumption
101
1990, pp. 417–420). But Abel (1999, p. 48) has shown that many of these shopping women who visited the Nickelodeons with their children were actually middle-class women.4 The price of cinema was probably an important factor for the kind of audience it interested. Before the Nickelodeon, prices varied from a dollar or more for the first special Lumie`re events, to a few cents to 50 cents for a travelling showman (Musser, 1990, p. 299). But in general, the market was in too turbulent a condition to put a reliable average price on motionpicture watching. This was even harder because they were often part of live entertainment. The prices the Nickelodeon charged were between 5 and 10 cents, which often enabled the spectators to stay as long as they liked. Around 1910, when larger cinemas emerged on key city-centre locations, more closely resembling theatres than the small and shabby Nickelodeons, prices increased. When the feature film had established itself as standard in about 1917, the average price was around 20 cents (Koszarski, 1990, pp. 13–15). However, substantial differences in prices existed. In individual theatres different seats often had different prices. Moreover, in the larger cities, prices were differentiated among theatres, with the city-centre theatres which showed the first run of films sometimes charging $1 to $1.50 for a performance, while the small and shabby neighbourhood cinema might still charge 5 cents for a sixth run. In between these two were stratifications of other theatres with different prices.5 The above indicates that a time lag existed of at least 12 years between the availability of the stable innovation and the take-off of cinema in 1907. This suggests that the null hypothesis can be rejected that cinema was nearly exclusively technology-driven and supply-led. During the 12-year lag, demand for entertainment grew steadily and people had more discretionary income left to spend on cinema, as will be discussed in the section below.
3. THE EVOLUTION OF ENTERTAINMENT CONSUMPTION 3.1. Total Consumer Expenditure Between about 1900 and 1940 over-all per capita expenditure on spectator entertainment showed a roughly similar long-run growth pattern in the US, Britain and France (Fig. 2). The average growth rates, although not having
102
GERBEN BAKKER
Real expenditure per capita (1914 = 100)
220 200 180 160 140 120
US
100 80
UK
60 40
FR
20 0 1880 1885 1890 1895 1900 1905 1910 1915 1920 1925 1930 1935
Fig. 2. Real Entertainment Expenditure Per Capita, US, Britain and France, 1881–1938 (1914 ¼ 100). Note: From 1900 onwards, the UK index includes admissions to sports matches (see Fig. 3). Source: Bakker (2001b); based on US Department of Commerce (1975), Prest (1954), Stone (1966), Cine´matographie Franc- aise, 1930, 1935, Durand (1956, p. 213).
entirely identical periods, were within a narrow range of 2.3 and 2.7 per cent per annum (Table 2).6 The 2.5 per cent per capita growth rate for the UK, compares to an average annual growth of real wages in industry of 1.0 per cent between 1881 and 1913, and 3.0 per cent between 1914 and 1938, or about 1.9 per cent for 1881–1938.7 Entertainment was a luxury, the consumption of which, in monetary terms, increased faster than real wages. The falling price of a spectator-hour of entertainment made the difference even higher in quantity terms. In the short-run, however, substantial differences existed. During the First World War entertainment expenditure moved in opposite directions in France and Britain and remained stable in the US. During the great depression US real entertainment expenditure shrunk substantially, while European levels remained stable. The French expenditure level was substantially lower than in the other two countries, about a fifth in 1938 using exchange rates, although the difference is difficult to quantify because of devaluation of the franc and purchasing power parity (PPP) issues. French expenditure also fluctuated more in the short term.
103
The Evolution of Entertainment Consumption
Table 2.
Average Annual Growth of Real Entertainment Expenditure, US, Britain and France, 1881–1938. US
UK
FR
Cinema and live 1881–1938 1900–1938 1909–1938 1914–1938 1934–1938
2.50 2.70 2.29 2.63 0.33 Cinema
1881–1938 1909–1938 1914–1938 1934–1938
10.99 8.06 1.24 Live
1881–1938 1900–1938 1909–1938 1914–1938 1934–1938
0.82 0.02 3.83 1.29 1.38
Source: Bakker (2001b, 2004).
The relative similarity of overall entertainment expenditure hid sharp differences in its composition. In the early 1910s, the expenditure share of live entertainment was roughly the same in the US as in France, but subsequently the US product mix changed sharply, with the share of live declining until the early 1920s (Fig. 3). From then on the difference in expenditure share remained stable. When sound film arrived (in 1927–1929), it declined in both the countries at about the same rate. In expenditure terms sound film made a similar relative impact in France as in the US, although price and quantity data would be needed to test this. The sparse UK data suggest the expenditure mix was roughly the same as in France (though the quantity mix was rather different, data for 1938, below, will show). Expenditure data for the US show a mild decline in net total expenditure between 1909 and 1921. This was composed of falling live expenditure and rapidly growing film expenditure. It is likely that the other countries experienced a similar substitution. Between about 1923 and 1925, US
Live entertainment (% of all spectator entertainment expenditure)
104
GERBEN BAKKER
100 US 80
FR
60
40 UK 20
0 1909
1914
1919
1924
1929
1934
1939
1944
1949
Fig. 3. Share of Live Entertainment Expenditure in Total Spectator Entertainment Expenditure, US, Britain and France, 1909–1951 (in per cent). Note: The British data includes admissions to sports matches, but could not be disaggregated further. For the tax year 1937–1938, it was estimated that sports admissions accounted for about 20 per cent of all non-cinema admissions (Stone, 1966, p. 81), and probably for far less of expenditure. Therefore, to estimate the British data, for all years the ticket price for sports matches is set at half the price of live entertainment, which results in the live expenditure share declining by between 2.3 to 2.4 percentage points. Source: Bakker (2001b); see sources in Fig. 2.
expenditure on cinema stabilised and live expenditure rebounded. Then, with the arrival of sound, cinema expenditure grew rapidly again and live expenditure fell sharply – well before the great depression started, showing that initially it was driven by sound, not depression. During the early depression years, cinema expenditure continues to grow, probably because sound film was still a novelty and substantially cheaper than live alternatives. Unemployment decreased both the opportunity costs of entertainment activities for many consumers and consumers’ purchasing power. People thus were encouraged to substitute even more cinema for live entertainment. After the First World War, expenditure on live entertainment always remained several factors lower than that on cinema, despite the rebounds in the 1920s and 1940s. Those rebounds might have been due to the recovery from economic recessions.
The Evolution of Entertainment Consumption
105
The differences between France and the US, and possibly also between Britain and the US, might be explained by the US dominance of European cinema screens from the late 1910s onwards (Bakker, 2005a, 2005b). This gave British and French live entertainment a competitive edge over cinema that the American live entertainment lacked. Before the coming of sound, the French live-entertainment industry offered consumers entertainment in the local social, cultural, political and intellectual environment. After sound, live entertainment gained a second competitive advantage because it was spoken originally in the local language.8
3.2. Early Consumer Surveys Few quantitative indicators exist on the demand for, and consumption of, entertainment. For household expenditure, and entertainment as a part of it, only some anecdotal, sparse, case-by-case data exist before the late nineteenth century. From the mid-nineteenth century onwards studies of the conditions of the working classes became more common, many inspired by the pioneering work of Fre´de´ric Le Play (1877). These early studies on family budgets seldom looked at expenditure on entertainment and recreation.9 The earliest scientific information is from Dorothy Brady (1972), who constructed representative sample budgets for American families in the 1830s, which are slightly above the relevant averages for each of three types of residential location: farm, village and city. Brady found relatively high expenditures on reading and recreation: about 2 per cent of all expenditures for all groups (Table 3). Church and charity outlays were even higher, varying from 9 per cent on farms to 3 per cent in cities. Possibly these items were over-reported, because giving generously could be considered socially desirable. Part of charity expenditure may also have been used like presentday social security contributions, especially in the farm and village communities. Farm families spent more on tobacco than they spent on reading and recreation, while city dwellers spent only 0.8 per cent and village families were caught in the middle.
3.3. The 1889–1890 Household Expenditure Survey Only in 1889–1890 was the first systematic household expenditure survey conducted, with a large number of respondents, and a sample that at least
106
GERBEN BAKKER
Table 3.
Estimated Breakdown of American Family Expenditure, 1830s.
Item
Farms
Villages
Cities
Purchased food Clothing House operation House furnishings Transportation Personal care Medical care Tobacco Reading, recreation Church, charity Not itemised Total (excluding shelter)
39.0 27.0 1.5 8.0 4.5 3.0 2.5 2.5 2.0 9.0 1.0 100.0
49.0 20.0 3.0 6.0 4.8 1.6 2.6 1.6 1.8 4.6 5.0 100.0
43.6 20.0 2.3 6.7 4.1 1.8 2.4 0.8 1.8 3.1 13.4 100.0
Source: Brady (1972, pp. 73, 76, 78; quoted in David & Solar, 1977, p. 41).
partially started to resemble a random sample. Under supervision of Carroll D. Wright, the US Commissioner of Labour, the US Department of Labour carried out a family expenditure survey, as part of a production cost study on nine protected industries (bar iron, pig iron, steel, bituminous coal, coke, iron ore, cotton textiles, woollens and glass).10 The survey is not fully random or representative because it selected and interviewed only workers in co-operating firms, because it selected only co-operating workers who provided information in sufficient detail, and because only industrial workers with families were included. Nevertheless, Michael Haines has shown that, at least for the United States, comparison with the US census gives some support to the representativeness of the data (Haines, 1979, pp. 292–295). The survey lists several categories relevant to leisure: expenditure on amusements and vacation, reading, liquor, religion and charity. The category ‘amusements and vacation’ includes live entertainment, but it is impossible to say which share went to sports matches, music hall, concerts or theatre. The survey showed that the US had the highest average income, and also the highest range of incomes, suggesting a more skewed income distribution than Britain and France (Table 4). The average household income was $684, substantially above the non-farm average in the U.S. at the time, which was $471 in 1889 and $475 in 1890. Probably more members of the sample households were working. Despite higher income and similar household
108
GERBEN BAKKER
size, average US entertainment expenditure was less than half that of Britain and France, and relative to income less than a third of that in Britain and France (1.11 per cent vs. 3.50 and 3.88 per cent). If the samples are broadly representative this suggests a sharp contrast in US and European consumer preferences. The income elasticity of entertainment expenditure can offer further insight into this difference. Given that many households did not report any entertainment expenditure, calculating income elasticity based on the whole sample may be misleading. We therefore use two alternative ways. First of all, we fit an ordinary least-square (OLS) model to the whole sample and calculate the (biased) elasticity accordingly. Second, we fit a logit model for the whole sample, investigating how the likelihood of positive entertainment expenditure increases with income, and then calculate income elasticity for the cases with non-zero expenditure in two ways: using an OLS model or a log–log constant elasticity model.11 Using the first method, income elasticity was substantially above unity, for all three countries suggesting that entertainment was a luxury everywhere (Table 5). In both the US and the UK, entertainment was highly income elastic, as elasticity was more than twice unity; while in France entertainment was substantially less elastic and far closer to being a normal good that was part of the necessities of the everyday French consumer in 1890. This is corroborated by the logit model (Table 5 and Fig. 4), which shows that at the lowest income levels, French households were three times as likely to have positive entertainment expenditure than US ones and two times as likely than UK ones. Consequently, the marginal effect of an additional dollar of income on the likelihood of positive French entertainment expenditure (0.03 per cent) was substantially smaller than elsewhere. The UK had the strongest marginal effect (2.5 times the French effect), while the US had the largest likelihood elasticity, a percentage increase in likelihood for a percentage increase in income. The latter is probably due to the US incomes being more dispersed to the right, yielding a larger average point elasticity. The income elasticities for households with non-zero expenditures are closer together than those for the whole sample, and also closer to unity. Also, entertainment was more of a luxury in the UK than in the US. A log– log constant elasticity model magnifies these differences, by bringing French and US elasticities close to unity, while UK elasticity remains far above unity. The category ‘amusements & vacations’ of the 1889–1890 survey, is, of course, an imperfect proxy, as it also contains expenditure for vacations. Another survey on Britain between 1891 and 1894 gives some ballpark
The Evolution of Entertainment Consumption
109
Table 5. Household Entertainment Expenditure in the US, Britain and France, 1889–1890.
OLS whole sample Average income ($) Average expenditure ($) Expenditure share (%) Coefficient of variation Income elasticity Impact income elasticity Logit model Share nonzero expenditure (%) Intercept (%) Marginal effect (%) Units to 100% ($) (100% likelihood)/average Point elasticity Impact effect Unit elasticity OLS for cases expenditure >0 Income elasticity Impact income elasticity Log-log for cases expenditure >0 Income elasticity Impact income elasticity
US
UK
FR
684 7.57 1.11 2.81 2.14 1.56
532 18.65 3.51 1.37 2.16 2.16
409 15.86 3.88 1.34 1.41 1.41
49 20 0.050 3236 4.7 0.71 15.0 0.63
71 29 0.076 1639 3.1 0.57 13.6 0.43
79 65 0.030 1444 5.5 0.16 6.2 0.13
1.43 1.43
1.68 1.68
1.26 1.26
1.13 1.16
1.80 3.51
1.07 2.79
Notes: For the respective impact effects, one standard deviation of income has been taken as the unit for each country. Most findings are significant at the 1% level, some at the 0.1% level. The heading ‘Entertainment’ refers to the survey item ‘Amusements and vacations’. Source: Data US Commissioner of Labor Survey 1891, provided by Michael Haines.
indication about the relative share of vacations and amusements. The Economic Club (1896) carried out a survey among 28 ‘industrial families’. The representativeness of the sample cannot be established, and the survey only recorded expenditure, not income. Average annual income for the 28 families was £92.16, or $449, considerably below the 1889–1890 survey average income of $532. Likewise the percentage of households with positive expenditure on ‘recreation’ and ‘travelling’ was lower, 54 per cent and 21 per cent, versus 71 per cent. The logit model predicts a value of 68 per cent for UK income of $449, suggesting that not all of the difference can be explained by the lower average income in the 1891–1894 sample. The total expenditure on ‘recreation’ and ‘travelling’ was 2.41 per cent, well below the 3.51 per cent of the 1889–1890 survey. Expenditure on recreation was
110
GERBEN BAKKER
Likelihood of positive expenditure on amusements and vacations (%)
100 90 80 FR 70 60 50 UK 40 US
30 20 0
Fig. 4.
500
1000 1500 Income ($)
2000
2500
Likelihood of Positive Expenditure of Households on Amusements and Vacations, US, Britain and France, 1889–1890.
roughly twice that on travelling: 1.62 versus 0.79 per cent. The income elasticity for aggregate expenditure on these two items was 1.63, considerably below the 2.16 elasticity estimated by OLS for the 1889–1890 survey, though not far from the OLS elasticity for cases with positive expenditure (1.68) and the log-log constant elasticity estimate (1.80). This elasticity breaks down into an elasticity of 1.31 for recreation and 2.29 for travelling, suggesting that, for the 1889–1890 survey, the elasticity for ‘amusements’ expenditure without ‘vacations’ may be substantially lower than the aggregate elasticity. The question remains how entertainment compared to expenditure on other leisure goods/services. For the US, logistic curves have been estimated for all leisure goods/services in the survey over the meaningful income interval of $0–$2,000 (Fig. 5). Three patterns attract the attention. First, a marked difference is apparent between liquor and tobacco on the one hand, and the other items on the other. The number of households that spend on liquor and tobacco was quite stable over the income interval, with liquor starting from quite a low initial value and rising slightly, and tobacco starting from the highest value in the group and declining slightly. The other four items rose quite substantially with income. Second, entertainment expenditure had the lowest starting value, with only a fifth of the families
111
The Evolution of Entertainment Consumption
Likelihood of positive expenditure (%)
100
Reading Religion
90 80
Tobacco
70 60
Charity
50
Liquor
40 30 Amusements/ Vacations
20 10 0 0
500
1,000
1,500
2,000
2,500
3,000
3,500
4,000
4,500
Income ($)
Fig. 5.
Likelihood of Positive Expenditure of Households on Various Leisure Goods, US, 1889–1890.
reporting expenditure, but rose the most rapidly with income. Although cross-section results cannot predict long-term changes, this nevertheless suggests that a rise in income could result in a disproportionately large rise in entertainment expenditure. Third, all the four other leisure goods/services approached 100 per cent as income approached the highest values (if we forget about the few outliers above $2,000), with all above 90 per cent at an income of $2,000. The comparison of expenditure across countries shows some marked differences (Table 6). As a share of income, French households spend about half the amount on reading as British and US households, British households spend about half as much on religion as their US counterparts, and French households about a third as much on religion as the British. The French also spend a fourth to a seventh the amount on charity as Britain and the US, and double or triple the amount on liquor. On leisure in total as a percentage of income, the French spent the most, followed by the British and only then the US households. In absolute (dollar) terms, however, the expenditure was roughly the same. The question remains to what extent these expenditures are comparable. The researchers in 1890 seem to have used the exchange rate to convert all
Descriptive Statistics of Income, Size and Entertainment Expenditure for Households, US, UK and France, 1889–1890. Income ($)
Sample size Mean Standard deviation Coefficient of variation Minimum Maximum Range Interquartile range Median Mode Mean expenditure/income (%)
Household Size (Persons)
Entertainment Expenditure ($)
US
UK
FR
US
UK
FR
US
UK
FR
6,809 684 337 0.49 84 4,500 4,416 378 597 600
1,024 532 235 0.44 177 1,582 1,405 286 462 389
263 409 238 0.58 43 1,737 1,694 220 347 290
6,809 4.78 2.12 0.44 1 15 14 3 4 4
1,024 4.95 1.94 0.39 1 13 12 3 5 5
263 4.99 2.15 0.43 1 12 11 3 5 5
6,809 7.57 21.25 2.81 0 600 600 6.00 0.00 0.00 1.11
1,024 18.65 25.46 1.37 0 204 204.39 24.30 9.70 0.00 3.50
263 15.86 21.23 1.34 0 193 193 18.23 9.70 0.00 3.88
The Evolution of Entertainment Consumption
Table 4.
Note: The heading ‘Entertainment’ refers to the survey item ‘Amusements and vacations’. Source: Data US Commissioner of Labor Survey 1891, provided by Michael Haines.
107
112
Table 6. Household Expenditure on Leisure Goods/Services, US, UK and France, 1889–1890. Expenditure (% of Income)
Amusements/vacations Reading Religion Charity Liquor Tobacco Total
Expenditure ($)
Coefficient of Variation
US
UK
FR
US
UK
FR
US
UK
FR
1.10 0.80 0.97 0.40 1.80 1.30 6.37
3.47 0.90 0.54 0.26 2.36 1.13 8.66
3.85 0.46 0.18 0.06 5.16 0.99 10.70
7.53 5.47 6.64 2.74 12.31 8.89 43.58
18.46 4.79 2.87 1.38 12.56 6.01 46.08
15.75 1.88 0.74 0.25 21.11 4.05 43.77
2.82 1.19 1.42 2.43 2.39 0.92 1.86
1.38 0.93 1.65 2.36 1.41 1.11 1.47
1.35 1.36 3.02 6.09 1.21 1.18 2.37
Range (max/min)
2.45 2.91 9.01 11.15 1.71 2.20 1.06
Note: Total for coefficient of variation is unweighted average. Source: Data US Commissioner of Labor Survey 1891, provided by Michael Haines.
GERBEN BAKKER
113
The Evolution of Entertainment Consumption
amounts into dollars. The exchange rate however, generally reflects international trade in goods and services and capital flows, although in 1890 most exchange rates were fixed by the gold standard. Since these leisure goods and services were partially untraded, PPP-ratios may be needed to get a more accurate reflection, but even then one could debate whether the comparison is meaningful or not. As would be expected, the expenditure range is the lowest for the two traded goods (liquor and tobacco) and far higher for the four non-traded services. When the expenditure on amusements/vacations is compared across the three countries against income, it is clear that in all three countries it was a luxury, with relative expenditure increasing as income increases, although it is less so in France (Fig. 6). Second, we see again the important difference between Europe and the US, now not only in levels, but also in the speed of increasing entertainment expenditure as income increased. US households spent less on entertainment, and expenditure rose less rapidly when income rose. Potential explanations could be the low relative price of entertainment in the US (so that in quantity terms the difference would be smaller), a lower US consumer preference for entertainment, or, on the contrary, that the relative price of entertainment was so high, because of the scarcity of skilled labour, that households could not afford to spend much on it. Data on
Expenditure amusements /vacations (% of income)
6 5 4 3 FR 2 UK
1
US 0 0.00
0.50
1.00 1.50 2.00 2.50 3.00 3.50 Income/household member (times average income)
4.00
4.50
Fig. 6. Expenditure on Amusements and Vacations Across Different Income Classes, for the US, Britain and France, in Times Average Income, 1889–1890.
114
GERBEN BAKKER
spectator entertainment in 1900 show that entertainment prices (compared at exchange rates) were high in the US and that the quantity consumed per capita was small, about a fifth compared to Britain (Bakker, 2004a). The quantity in France was tiny and prices were very high. Given the low per capita expenditures in 1900, however, it is questionable whether the two data sets are comparable. 3.4. The Surveys in the 1930s Similar surveys were carried out in the US between 1934 and 193612 (and also in 1917-1919),13 in France between 1936 and 1938,14 and in Britain between 1937 and 1939. Although the years differ, the data are probably quite comparable. The US data appear to be the most reliable, because it has a reasonable standard deviation compared to income, and a low coefficient of variation. The other samples have large standard deviations and coefficients of variation, which raise questions on their representativeness (Table 7). Differences in income distribution may not be the cause as British inequality was similar to the US, the top quintile receiving half of national income.15 For France, where just 92 households in one city were surveyed, the data can only be taken as a first rough indication of spending patterns. The French dollar income is over an order of magnitude lower than British income and over two orders of magnitude lower than US incomes, obviously the result of the devaluation of the French franc. PPP-data would probably result in different relative incomes. In the US, the income share of entertainment expenditure increased from 0.70 in 1918 to 1.22 per cent in 1935. The income elasticity fell from 2.54 to 1.02, making entertainment nearly a normal good, and not incompatible with the fact that, at the cross-section level, as income tends to infinity, income elasticity of a specific item eventually approaches unity, if expenditure increases linearly with income.16 For all three countries, income elasticities were roughly in the same ballpark in the 1930s, and not very much different from unity. For cinema expenditure, we see a similar drop in US elasticity, while a marked transatlantic difference surfaces: in the US cinema is a luxury, with expenditure rising faster than income, in Europe cinema is inferior. Within Europe, cinema was most inferior in Britain. A decline in the luxuriousness of entertainment expenditure relative to the 1890s concurs with Owen’s findings that the aggregated expenditure on leisure and recreation as a percentage of GDP increased substantially between 1900 and 1930, but remained stable after that, at about 5 per cent of GDP, at least until the late 1960s (Owen, 1970, pp. 86–94).
Descriptive Statistics of Income, Size and Entertainment Expenditure for Households, US, UK and France, 1918, 1934–1939. Income ($ of 1936)
Sample size (households) Mean Standard deviation Coefficient of variation Minimum Maximum Range Interquartile range Median Mode Mean expenditure/ income (%) Income elasticity (mean maximum)
Household Size (Persons)
US18
US35
UK
FR
12,096
14,496
3,580
92
1,391 377
1,543 265
1,560 971
69 32
0.27
0.17
0.62
0.47
748 2,566 1,818
979 2,426 1,447
430 5,308 4,878
42 145 103
US18
Entertainment Expenditure ($)
US35 UK FR US18 US35
UK
FR
Cinema Expenditure ($)
US18 US35
UK
Live Entertainment Expenditure ($)
FR US18 US35
UK
FR
4.9
3.6
3.6
3.7
9.74
18.83
24.49
2.62
6.82
16.98
12.64 1.59
0.83
0.46
7.33
0.43
4.3 6.4 2.1
2.0 6.5 4.5
2.8 3.8 1.0
3.1 3.8 0.7
2.54 34.89 32.35
7.64 25.72 18.08
4.04 61.04 57.00
1.58 5.65 4.08
1.87 22.58 20.71
6.17 29.84 23.67
3.09 1.07 22.29 3.19 19.20 2.12
0.07 4.87 4.80
0.010 0.69 3.64 28.66 3.54 27.97
0.19 1.19 1.00
0.70
1.22
1.57
3.81
0.49
1.10
0.81 2.31
0.06
0.03
0.047 0.63
2.54
1.02
0.74
1.04
2.32
1.49
0.49 0.89
4.40
8.16
1.22
The Evolution of Entertainment Consumption
Table 7.
1.54
Note: UK data concerns expenditure, not income.
115
116
GERBEN BAKKER
Income elasticity for live entertainment expenditure was far larger in the US than in Europe, and it even doubled between 1918 and 1935. The main reason seems to be the small and declining share in income; the smaller the expenditure item, the larger income elasticity, if we assume expenditure increases linearly with income. The decline of live entertainment expenditure was largely caused by cinema, and especially by sound films. The large volume-low margin-high profit part of live entertainment was automated away by cinema, while what remained split into a highly commercial/profitable metropolitan low-volume-high margin-high profit part and a heavily subsidised low volume-low margin-low profit part (Bakker, 2004). In Britain and France the relative live expenditure was far higher, suggesting that cinema was less of perfect substitute (for example, because most films shown were not in the consumers’ mother tongue) or that those countries had more competitive live entertainment industries that had a comparative advantage (lower relative price) compared to the US.17 For all countries, the entertainment income elasticities were lower than the amusements and vacations elasticities in 1890 (not a perfect comparison, as vacations are absent in the 1930s). Again, as total income increases, it could be expected that income elasticity declines, although this concerns times series rather than cross-sections. Only for the US, for 1918, was the income elasticity substantially higher than in 1890. When the shape of entertainment expenditure relative to income is examined (Fig. 7), a similar order as in 1890 shows: expenditure was highest in France and lowest in the US, roughly across all income classes. The distribution of US and French income was less dispersed than British income, which contained extremes for both low and high incomes. Britain and France also had less steep curves than the US. Further, a marked increase in US expenditure was visible between 1918 and 1935, consistently for all income groups, although the shape got slightly steeper. For cinema expenditure (Fig. 8) the pattern changed, and US expenditure overtook British expenditure at about 0.80 of average income. Even 1918 US expenditure overtook British expenditure in the last-income class; assuming the British curve would have been lower in 1918 as well, this suggests a similar US–UK pattern for 1918. France showed a sharp drop in expenditure from the first to the second-income class, and then a slightly increasing curve. For live entertainment (Fig. 9), the order of magnitude difference between the US and Europe clearly showed, as well as a far slower increase with income in the US than in Europe. British and French expenditures were quite close and exhibited broadly similar patterns.
117
The Evolution of Entertainment Consumption
Entertainment expenditure (% of income)
4.5 4 3.5
FR
3 2.5 2 1.5 UK
1
US1935
0.5
US1918
0 0.20
Fig. 7.
0.60
1.00 1.40 1.80 2.20 2.60 3.00 Income/household member (times average income)
Entertainment Expenditure Across Income Groups, in Share of Average Income, US, Britain and France, later 1930s.
2.50 Cinema expenditure (% of income)
3.40
FR
2.00
1.50
1.00 US1935
UK 0.50
US1918 0.00 0.20
Fig. 8.
0.60
1.00 1.40 1.80 2.20 2.60 Income/household (times average income)
3.00
3.40
Cinema Expenditure Across Income Groups, in Share of Average Income, US, Britain and France, later 1930s.
Live entertainment expenditure (% of income)
118
GERBEN BAKKER
0.80 0.70 0.60 0.50
FR
0.40 0.30 0.20 UK
0.10
US1918
0.00 0.20
Fig. 9.
US1935 0.60
1.00 1.40 1.80 2.20 2.60 Income/household (times average income)
3.00
3.40
Live Entertainment Expenditure Across Income Groups, in Share of Average Income, US, Britain and France, later 1930s.
3.5. Decomposing International Consumption Differences in Technology and Taste Effects Price and consumption per capita data for 1938 (from Bakker, 2004) enables the calculation of national budget constraints (Fig. 10).18 When a country can potentially buy more cinema tickets, it can also buy more live entertainment tickets. It is also evident that, while US and French ‘technology’ (relative prices) are broadly similar, Britain had a low price for live entertainment (Table 8) and its share was as much as 25 per cent, compared to just over 2 per cent in the US and 10 per cent in France. This difference must have been at least partially due to a different organisation of entertainment production rather than exclusively to consumer preferences. It is possible to formally decompose these national differences in consumption patterns into those due to differences in relative price (‘technology’) and those due to differences in consumer preferences (‘taste’). First, it is assumed that consumers spend a constant income share on spectator entertainment, and then divide it between cinema and live entertainment. Second, it is assumed that the relative price pc/pl (the slopes of the budget constraints in Fig. 10) largely reflects differences in production technologies rather than differences in demand.
The Evolution of Entertainment Consumption
119
90 80
Quantity of cinema (spectator-hours)
70 60 US
UK
50 40 30 FR
20 10 0 0
10
20
30
40
50
60
70
80
90
Quantity of live entertainment (spectator-hours)
Fig. 10. Live vs. Entertainment Quantities Consumed and Budget Constraints, Average per Capita, US, Britain and France, 1938. Source: Corrected estimates from Bakker (2004).
Given pc/pl, consumption preferences can be characterised by the quantity elasticity of substitution eqs (which defines the position of the data point on the budget constraint). The latter is the percentage change in cinema-hours for a percentage change in live-hours. Consumers chose a certain ‘exchange rate’, a certain value of eqs, which is defined as follows: qs ¼
%Dql qc dql qc 1 a ¼ ¼ ¼ 1a %Dqc ql dqc ql TRS
(1)
where qc is the amount of spectator-hours of cinema consumed, ql the spectator-hours of live entertainment, TRS the technical rate of substitution and a the share of cinema expenditure in total expenditure on live entertainment and cinema.
120
GERBEN BAKKER
Table 8.
US UK FR
Indicators of the Consumption of Live and Cinema SpectatorHours, Britain, France and the US, 1938.
qc+ql
qc
ql
a
g
pc/p1
eqs
54.2 76.5 14.7
53.0 57.1 13.2
1.2 19.4 1.5
0.92 0.67 0.68
0.978 0.75 0.898
0.26 0.68 0.25
11.34 1.99 2.17
100 263 96
100 18 19
Index (US ¼ 100) US UK FR
100 141 27
98 105 24
2 36 3
100 72 74
100 76 92
qc ¼ the number of spectator-hours of cinema consumed; ql ¼ the number of spectator-hours of live entertainment consumed; a ¼ the expenditure share of cinema consumption; g ¼ the quantity share of cinema consumption; pc/pl ¼ the relative price of cinema over live entertainment; and eqs ¼ the quantity elasticity of substitution of cinema for live entertainment. Note: All figures are national averages per capita for 1938. Source: Corrected estimates from Bakker (2004).
It is clear that ‘consumer preferences’, as proxied by this quantity elasticity of substitution of cinema for live entertainment, were not the same across the countries. The US had an incredibly high eqs, meaning that the US consumer, by reducing the quantity of cinema consumed by 1 per cent, could increase the quantity of live entertainment consumed by 11 per cent. In France, eqs was 2.2, and in Britain it was only 2.0 (Table 8). This suggests that besides technology, consumer preferences were also important to explain national differences. The difference between countries in the quantity share (g) of cinema in all spectator entertainment consumed can be decomposed into the effects of eqs (‘taste’), differences in pc/pl (‘technology’) and their joint effect. Their relative magnitude is calculated using the equations below. The intercepts and points on the budget constraints are defined as follows: 1 þ qs qc qs
(2)
qs qmax 1 þ qs c
(3)
qmax ¼ ð1 þ qs Þql l
(4)
qmax ¼ c
qc ¼
The Evolution of Entertainment Consumption
ql ¼
121
1 qmax 1 þ qs l
(5)
qc qc þ ql
(6)
Yielding: g¼
1 qmax 1 þ qs l g1 Dg ¼ g2 g1 ¼ qs 1 max q qmax þ 1 þ qs l 1 þ qs c qmax 1 l ¼ g1 1 þ qs qmax þ qs qmax c l
ð7Þ
where qmax and qmax are the maximum amounts of cinema and live enterc l tainment that can be consumed. To examine the effect of technology, tastes are kept constant; qmax is l computed using the relative price (slope) of the comparator country (and keeping all other variables constant), and introduced as qmax in (7), giving l2 the result: ! max max ðq q Þ qs qmax l l c 2 1 Dg ¼ (8) max 1 þ qs ðqmax þ qs qmax þ qs qmax c Þðql 1 c Þ l2 To examine the effect of tastes, technology is kept constant; the qs of the comparator country is introduced in (7) as qs2 (keeping all other variables constant), giving the result:19 1 1 Dg ¼ qmax (9) l ð1 þ qs2 Þðqmax ð1 þ qs1 Þðqmax þ qs2 qmax þ qs1 qmax c Þ c Þ l l To measure the effect of technology, for example, the US relative price has been set at the UK relative price, keeping elasticity constant (Table 9). The effects can be measured in two directions and the average effect, which cancels out the joint effect, gives a rough and ready estimate of the relative importance of technology and taste in explaining country differences. Although not explicitly assumed, this method is consistent with Cobb– Douglas preferences and Cobb–Douglas technology/taste decomposition gives exactly the same results.20 Thus the difference in relative cinema consumption between the US and Britain can be explained for about three-fifths by technology and for about two-fifths by taste. Given that the data are not extremely precise, this
122
GERBEN BAKKER
Table 9. The Effect of Relative Price and Quantity Elasticity of Substitution on Differences in Cinema Consumption, US, Britain and France, 1938. In Percentage-Points Effects
US to UK UK to US Average US to FR FR to US Average UK to FR FR to UK Average
Total difference % points
dpc/pl (% points)
23.2 23.2
3.4 13.9 8.7 0.1 0.4 0.2 14.4 13.6 14.0
8.0 8.0 15.2 15.2
In Percentage of Total Difference Effects
de (% Joint (% points) points) 9.2 19.8 14.5 8.4 8.1 8.2 1.6 0.8 1.2
10.5 10.5 10.5 0.3 0.3 0.3 0.8 0.8 0.8
Total difference %
dpc/pl (%)
de (%)
Joint (%)
100 100
15 60 37 1 5 3 95 89 92
40 85 63 105 101 103 11 5 8
45 45
100 100 100 100
4 4 0 5 5 0
Index (US to UK ¼ 100) 100 100 34 34 66 66
15 60 37 0 2 1 62 59 60
40 85 63 36 35 35 7 4 5
45 45 1 1 3 3
dpc/pl, the difference in relative price (‘technology’); and de, the difference in the quantity elasticity of substitution (‘taste’). Notes: All figures are national averages per capita for 1938. Average refers to the average size of the effect in absolute terms, not to the direction. Source: Appendix, Tables A1, A2 and A3; corrected estimates from Bakker (2004).
suggests that the lower price and differences in taste were about equally important for the large quantity of British live entertainment consumed.21 The difference between Britain and France, on the contrary, can be explained almost exclusively by differences in technology. The difference between France and the US, on the contrary, is far smaller and could be wholly explained differences in taste. These findings suggest that UK had a clear comparative advantage towards live entertainment, the US towards cinema, while the situation of France was
The Evolution of Entertainment Consumption
123
undetermined. Unfortunately for Britain, live entertainment could hardly be traded, meaning that a specialisation on live entertainment would yield less advantage to Britain than a specialisation on cinema yielded to the US.
4. COMPOSITION AND GROWTH OF RECREATION EXPENDITURE AS A WHOLE The last quarter-century has seen a number of scholarly attempts to introduce an evolutionary framework to the study of organizational ecology and economic change (Nelson & Winter, 1982; Hannan & Freeman, 1989). This section aims to widen the framework employed so far to help us choose between our two competing explanatory hypotheses by investigating the potential role of wider consumption routines, skills and capabilities in the evolution of demand through processes such as selection, replication, imitation and modification and through random events and ‘mutations’. The preferred hypothesis – that the emergence of cinema was mainly demand-led – is retained, but it is shown that it is compatible with an evolutionary account. Consumers started to spend more time and money on leisure activities, and initially their expenditure was spread out among many different categories. Much of the demand, however, went to spectator entertainment, and to reduce bottlenecks and increase revenues, entrepreneurs started to use cinema technology. Consumers reacted favourably to this technology, giving entrepreneurs incentives to develop it further. Over time, in an evolutionary process, more and more expenditure was moved away from things such as tobacco and alcohol to entertainment expenditure, and within entertainment expenditure, more and more was spent on cinema. Cinema-going became a habit for consumers, sometimes daily, sometimes weekly. The outcome of the evolutionary process was that cinema became the dominant form of entertainment. To sketch the environment in which demand for live entertainment and cinema boomed, the developments in other recreation products and services are outlined briefly and broadly. The rise of cinema took place within a general rise in demand for recreation products, influenced by five factors: more time, more money, urbanisation, new transport networks and population growth. The rise in expenditure on spectator entertainment was not simply a redistribution of existing recreation expenditure, but was also connected to these five underlying forces. In historical and cross-country research many types and kinds of quantitative information exist that are often difficult to compare because of
124
GERBEN BAKKER
differing time spans, units, accuracy of measurement and reporting. To overcome this, a method of informal comparative growth analysis and simulation is used, which allows us to compare the data systematically, by converting all-time series data into real per capita growth rates. This makes data more comparable across industries/markets and countries, and the combination of quantity and expenditure growth rates allow ballpark estimates of real and relative price growth rates, which otherwise would be difficult to obtain for many smaller goods and services. This method is far from perfect and not very precise, but it offers a better way of comparing different types of incomplete historical data than many alternative methods. It is no more than a rough and ready approach to get insight into the order of magnitudes of growth of leisure spending and on relative growth rates. Since data are for three countries and for differing time-spans, aggregate growth rates do not accurately reflect ‘real’ national growth rates. They are no more than abstract constructs that shed some limited light on relative growth rates in the absence of complete and fully comparable data series. One can also do simulation experiments with the rates allowing more insights into the workings of the process, not unlike simulations in evolutionary or population ecology and organisational ecology. For printed media, audiovisual media, sports, non-market entertainment and alcoholic drinks growth rates have been calculated from dispersed data sources that could be tracked down (Table 10). Except for drink, these rates suggest that the increasing demand for spectator entertainment was part of a broad-based boom in demand for recreation as a whole, both for commercially provided and non-commercial recreation. Although the consumption of all goods grew rapidly, audiovisual entertainment consumption grew about twice as fast as the average for all groups, while expenditure increased only slightly above average, suggesting a substantial decline in the real price. The substantial increase in quantity was hidden by an exceptional fall in price. This was at least partially brought about by substitution of films for live entertainment. Even compared to prices of other leisure goods, entertainment prices fell substantially. If the growth rates are hypothetically applied to the whole 1890–1938 period, audiovisual prices in 1938 were only 61 per cent of what they had been expressed in quantity of printed matter in 1890, only 74 per cent expressed in sports tickets and only 45 per cent expressed in drinks. Three audiovisual goods with high growth rates – the automated piano, cinema and radio licenses – brought up the average substantially. The growth intensity for sports suggest substantial scale effects, as recreation facilities attracted and could handle more and more consumers.
Category
Unweighted averages
Coefficient of variation
Printed
Product
Printed Audiovisual Sports Non-market Drink Group average All-item average Printed Audiovisual Sports Non-market Drink Group average All-item average Newspapers Periodicals Newspapers Newspapers Books and periodicals Published editions of adult fiction Published editions all classes of books Unweighted average Standard deviation Coefficient of variation Piano Automated piano
Country
Period
Growth per Capita per Annum
From
To
Quantity
1897 1898 1901 1913 1860 1894 1894 1850 1850 1875 1896 1832 1861
1937 1931 1923 1941 1903 1927 1927 1950 1940 1947 1941 1923 1940
2.53 10.85 4.55 7.76 0.57 5.25 5.78 0.21 1.01 0.96 0.64 1.98 0.96
US US UK UK UK UK UK
1850 1904 1900 1900 1900 1911 1911
1950 1947 1938 1919 1938 1935 1935
2.45 2.40
US, UK
1897 1850
1937 1950
US US
1850 1900
1909 1919
Expenditure
3.38 3.28 4.18 1.12 2.99 3.27 0.00 1.31 0.42 0.00 0.43
Intensity
6.50 6.07 6.29 6.40
Price after 48 Years
Price
Real Price
0.33 0.70 0.08
1.03 0.00 0.62
0.61 1.00 0.74
0.97 0.13 0.39
1.66 0.83
0.45 0.67
0.33
1.03
0.61
0.41 0.00 0.21
3.37 2.68 3.39 3.39 1.74 2.53 0.53 0.21 4.10 15.00
3.38 0.01 0.00
125
Audiovisual
Per Capita Growth of Leisure Goods and Services, US, Britain and France, 1832–1950, Real Expenditure and Intensity.
The Evolution of Entertainment Consumption
Table 10.
Category
Piano Piano Phonograph rolls and discs Cinema and live entertainment Cinema and live entertainment Cinema and live entertainment Cinema and live entertainment Cinema and live entertainment Cinema and live entertainment Cinema and live entertainment Live entertainment Live entertainment Live entertainment Live entertainment Cinema Cinema Radio licenses Radio licenses Unweighted average Standard deviation Coefficient of variation Baseball attendance Baseball attendance Sports and games National football league revenue First division league football attendance First Division League Football attendance
Country
Period
Growth per Capita per Annum
From
To
UK FR UK US US US UK UK FR FR US UK UK FR US FR UK FR US, UK, FR
1870 1850 1905 1900 1909 1900 1900 1881 1900 1914 1909 1881 1900 1914 1909 1914 1923 1933 1898 1850
1910 1910 1930 1938 1938 1940 1938 1938 1938 1938 1938 1900 1938 1938 1940 1938 1938 1938 1931 1940
US US UK UK UK
1901 1921 1900 1937 1888
1921 1940 1919 1947 1913
UK
1888
1913
Quantity
Expenditure
Intensity
Price after 48 Years
Price
Real Price
0.66
0.04
0.98
0.03
0.73
0.70
0.56
0.14
0.94
0.70
0.00
1.00
2.10 2.25 10.60 5.59
2.63
2.29 2.79 2.72 2.50
5.99 2.63 3.83 1.92 0.02 1.29 10.99 8.06 32.00 28.00 10.85 10.92 1.01 3.21 0.01
3.28 4.30 1.31
5.95 2.41 10.45 6.67
GERBEN BAKKER
Sports
Product
126
Table 10. (Continued )
Drink
UK UK US, UK
1875 1897 1901 1875
1914 1914 1923 1947
US US US US US US US US US US
1896 1924 1924 1916 1915 1910 1904 1904 1923 1913 1896
1941 1941 1941 1941 1941 1941 1941 1941 1941 1941 1941
UK UK UK FR FR FR FR FR FR US, UK, FR
1870 1877 1877 1832 1872 1832 1872 1832 1872 1860 1832
1919 1912 1912 1872 1923 1872 1923 1872 1923 1903 1923
4.55 4.37 0.96 15.00
4.18 1.77 0.42
9.72 3.12 6.50 2.70 0.41
0.08
0.62
0.74
0.97
1.66
0.45
6.07 13.41 3.86 4.52 5.58 13.90 3.41 2.41 7.76 5.00 0.64
6.07 0.00 0.00 1.12
1.67 0.59 1.15 0.70 1.88 0.78 1.86 0.46 0.57 1.13 1.98
1.12 0.00 0.00
127
Intensity ¼ attendance/production per unit; e.g. number of spectators per stadium, users per bowling alley, etc.Real price ¼ the percentage per annum with which the price difference between the good in question and audiovisual good changes.Price after 48 years: this is the change in the hypothetical relative price of audiovisual goods compared to the good in question, applying the growth rates to 1890–1940; for example, in 1938, the relative price of audiovisual entertainment, expressed in printed matter, was 61 per cent of what it had been in 1890. Source: Bakker (2001b).
The Evolution of Entertainment Consumption
Nonmarket
Cup final average crowd Rugby league attendance cup final Unweighted average Standard deviation Coefficient of variation Membership bowling associations Members per bowling alley Softball diamonds per capita Bathing beaches Swimming pools Supervised playgrounds Visits to national parks Visits to forests Hunting licences Unweighted average Standard deviation Coefficient of variation Alcohol Spirits Beer Wine Wine Beer Beer Spirits Spirits Unweighted average Standard deviation Coefficient of variation
128
GERBEN BAKKER
A year-on-year growth of about 6 per cent in utilisation, even if we allow for additional capital to build larger venues, is phenomenal. It results in venues that in 1938 would be 16 times as large as those in 1890. Urbanisation and transport networks also contributed to increased utilisation. The scale effects probably also mitigated potential rises in prices, as average costs would come down substantially and continuously, until full venue size was reached. The growth of non-market goods is surprisingly fast, and shows that increased cinema consumption cannot be fully explained by consumers substituting informal, traditional, non-market recreation for commercialised entertainment. The rapid increase in leisure time, the efforts of states and communities to provide leisure goods and the low price of non-market recreation probably all played a role. The similar growth rate of audiovisual products and non-market recreation could be driven by strong public good characteristics. Both have a strongly non-diminishing character. An additional person watching a film does not deplete the copyright, and hardly depletes the celluloid, an additional person visiting a national park or a playground diminishes the resource only at a low rate. Both also have some non-excludability properties: only copyrights enabled strong
Radio UK
Growth rate per annum (%)
30 25
Quantity Expenditure Intensity Price
Radio FR
20 Bowling assoc. National parks Softball Recorded music Cinema US Football Cinema+live, US/FR Cinema FR Sport/games UK Playgrounds Bowling
Automated piano
15
Cup Final
10
1st div. football Piano US
5
Newspapers US
Hunting US
Spirits FR
0
1840
Piano FR
1850
1860
1870
1880
Football
Live UK
1890 1900 Spirits UK
1910
-5
1920
1930
Live FR Live US
1940
1950
Mid growth-interval year
Fig. 11. Calendar Growth Interval Midpoint vs. Annual Per Capita Growth Rate, Various Leisure Goods and Services, US, Britain and France, 1850–1950. Note: The average growth interval midpoint is 1911 (median of the intervals is 1919, standard deviation 20 years, and coefficient of variation 0.3). Source: Table 10.
129
The Evolution of Entertainment Consumption
4,000 60 3,000 40 Total quantity
Audiovisual
2,000
Non-market 20
Sports
1,000
Drink Printed 0 1890
Total recreation quantity (1890=100)
Share of total quantity of leisure consumption (%)
excludability for filmed entertainment, and radio and television were largely non-excludable. Likewise, it is not easy, although possible, to exclude people from national park, playgrounds or public softball diamonds. High fixed and sunk costs in both cases mean that average costs will decrease for a long interval – for films even when sales equalled the entire market – so that prices could be relative low in competitive situations or when a social planner wants to maximise total economic welfare. When the growth rates are plotted against time (Fig. 11), it becomes evident that the highest growth rates tended to take place later, although a substantial part of this can be explained by the late average of the growth intervals (1911). Second, radio was an outlier that significantly increased the audiovisual growth rate. Third, nearly all the growth rates above 5 per cent were due to either audiovisual or non-market entertainment, and again took place in later years. The above growth rates enable a simulation of growth patterns. If the five categories are assumed to have had equal shares of recreation quantity consumed in 1890, then it is possible to estimate hypothetically what those shares would have been in other years, using actual growth rates (Fig. 12).
0 1895
1900
1905
1910
1915
1920
1925
1930
1935
1890 + Length of growth interval (years)
Fig. 12. Hypothetical Share of Audiovisual Entertainment in Total Quantity of Leisure Goods/Services Consumed as a Function of Equal 1890 Shares, with Historically Observed Growth Rates Applied from 1890 Onwards, and Hypothetical Quantity Consumed 1890–1938 (Percentage Share and 1890 ¼ 100). Source: Table 10.
130
GERBEN BAKKER
Because of its high growth rate, audiovisual entertainment would have reached 50 per cent in 1914, when the feature film was becoming the industry standard, and would have reached about 75 per cent by 1938.22 Non-market goods were the only other category that would have kept a meaningful market share, peaking at 26 per cent in 1910, and remaining 19 per cent in 1938. The recreation market as a whole grew tremendously, reaching nearly forty times its 1890 size by 1938 (Fig. 12).23 One can also examine how the audiovisual market share by 1938 depends on its initial share in 1890 (Fig. 13). Even if audiovisual entertainment had an initial market share of only 2.5 per cent in 1890 (and the remaining 97.5
100%
Share of entertainment expenditure in leisure expenditure in 1938 (%)
18,000
90%
Non-market 80%
16,000
Sports 14,000
70%
Audiovisual
60%
12,000 10,000
50%
Total
40%
8,000 6,000
30%
Contribution 20%
4,000
10%
2,000
Printed
0% 0
10
20
30
40
50
60
70
80
90
Total leisure market (1890=100) and contribution to growth rate *1000 (percentage points)
20,000
Drink
0 100
Share of entertainment in leisure expenditure in 1890 (%)
Fig. 13. Hypothetical Share of Audiovisual Entertainment in Total Quantity of Leisure Goods/Services Consumed in 1938, as a Function of the 1890 Share. Note: Total (solid black line) ¼ total leisure market size in 1938, in quantity (1890 ¼ 100); the size of this market increases as the 1890 quantity share of audiovisual entertainment increases. Contribution to growth rate (dotted line) ¼ the additional growth rate relative to the hypothetical growth rate without audiovisual entertainment (which was 5.4 per cent per annum). A value of 4,000, for example, means that 4,000/1,000 ¼ 4 per cent-points additional growth was caused by audiovisual entertainment. Source: Table 10.
The Evolution of Entertainment Consumption
131
per cent was divided equally among other categories), then it would have reached a 26 per cent share by 1938. For 5 per cent this would have been 40 per cent, and for 10 per cent 58 per cent. This suggests that audiovisual entertainment made a significant growth contribution to recreation as a whole. It probably goes too far to say that the boom in demand for live entertainment directly forced the emergence of cinema discussed earlier in this paper. Nevertheless, without sharply rising demand, for a long time the cinematograph would have remained what it had been during its first years: a novelty, a specialty, a luxury product every now and then shown in theatres and schools, or occasionally by travelling showmen. Cinema would not have taken off as a large-scale industry.24 The huge growth burst of demand enabled cinema to develop into a specialised industry with its own dedicated distribution delivery system. Without this boom in demand, the market would probably have been too small for a separate distribution delivery system, and costs sunk in film productions would have had to remain limited, hampering the possibility of films to rapidly increase their audience. Prices of film performances would probably have remained closer to the prices of theatre and big-time vaudeville, further preventing any take-off of a new industry.
5. CONCLUSION The paper explored the hypothesis that the take-off and growth of the film industry was largely demand-driven. A time lag of about 10 years existed between the availability of cinema technology and the take-off of the industry. This lag was short, but long enough to allow rejection of the nullhypothesis that technology alone accounted for the development of the film industry. The alternative, or demand-driven explanation, was sustained through examinations of total consumer spending, cross-section studies of household expenditure, estimates of cross-country taste/technology differences and an informal comparison of growth rates of different recreation goods and services. These investigations into consumption do not provide sufficient reason to fully reject a technology-based account, but they do justify a strong presumption in favour of demand factors. Two stories, then, can be told about the emergence of the film industry. The most popular one so far has been the story about great men, genial inventors who invented all the necessary components for film technology
132
GERBEN BAKKER
step-by-step, starting with projection in the 1850s, celluloid in the 1860s, roll-films in the 1880s, and finally the cinematograph in the 1895. The heroes of this story are men such as George Eastman, Georges Marey, Louis Lumie`re, Thomas Edison, William Kennedy Dickson, Robert Paul, Friese Green and Albert E. Smith. According to this story, these great men with their great inventions laid the foundation of the modern twentieth-century film industry, and on their shoulders stood great men, great entrepreneurs that quickly marketed their innovations to the world, men such as Charles Pathe´, Le´on Gaumont, Charles Urban or Carl Laemmle. In the second story, however, the great men are just ripples on great waves, and while their great invention singularly may have highly contributed to scientific knowledge, it hardly contributed to the take-off of the film industry. Their invention might have stayed a mere gadget, a visual toy and a novelty, as was the fate of its predecessors in the nineteenth century. It might have remained a premium product for a limited elite, as was the fate of its major fellow traveller, the phonograph. It might have become a fad for a short spell, such as the skating rinks and bowling alleys of the 1900s. Instead, because of fundamental changes in the composition and growth of consumer demand, it broke out of the control of its great inventors and quickly developed into the greatest entertainment industry of all.
NOTES 1. For a partially quantitative case study on the rise of Nickelodeons in Manhattan, see Ben Singer, (1995). Singer updates an earlier work by Robert Allen (1979). 2. See also Gerben Bakker (2005b). 3. A. J. P. Taylor (1976, p. 181), for example, writes enthusiastically: ‘‘Women joined their husbands in enjoyment, as they had never done at football matches or other public pleasures’’. 4. Within film history, substantial research has been done into the composition of early cinema audiences, generally in a qualitative way. The current paper does not aim to analyse cinema audiences socially or culturally; it only provides some perspective in this section as a background to the quantitative analysis that will follow. Film historical works on audiences include Robert Sklar (1975); Thomas Elsaesser (1990, 1994); David Bordwell, Janet Staiger, and Kristin Thompson (1985); Douglas Gomery (1992); Miriam Hansen (1991); Steven Ross (1998); Janet Staiger (1990); John Sedgwick (2000); Jeffrey Richards (1994); Claude Forest (1995); Georges Sadoul (1962); Jean-Jacques Meusy (1995). 5. For detailed historical research on cinema prices in 1900s London, see Burrows (2004), Sedgwick (1998) contains a detailed case study of price differentiating in 1930s’ Britain.
The Evolution of Entertainment Consumption
133
6. The series are not entirely comparable, as the British one includes admissions to sports matches from 1900 onwards (see Fig. 3 and Stone, 1966, p. 81). 7. Mitchell (1993, pp. 182, 184), combined with Mitchell’s consumer prices deflator (pp. 847, 849). The two series could not be linked because they do not overlap. The two rates have therefore to be combined to form a 56-year period to calculate the approximate average annual growth. 8. Dubbing still yields a film of a different quality than an original language film, in which local actors directly speak the local language. Differentiation may also explain the surprising rebound of French live entertainment expenditure in the late 1940s, when it reached roughly the same level as expenditure on cinema – the explanation of which is not the purpose of this work. Because of the war, French film production was temporarily halted, and possibly the cinemas could not provide enough locally made entertainment to constitute a satisfying mix. 9. See, for example, also Horrell (1996, pp. 561–604). The many early nineteenth century family budget studies Horrell used do not contain information on entertainment expenditure. For an overview of numerous early family budget studies, starting as early as the middle ages, see Nystrom (1931) and Zimmerman (1936). 10. The author wishes to thank Michael Haines for generously making available the computerised data of the survey. This research is discussed in detail in Haines (1979, pp. 289–356). 11. Method based on Feinstein and Thomas (2002). 12. Researchers interviewed 14,469 ‘‘white and negro’’ families from 42 cities spread over the US, each with a population of over 50,000 inhabitants. Families included in the survey had an income of at least $500 a year, received no relief during the survey year, had at least one earner employed for 36 weeks and earning at least $300. No clerical workers earning over $200 a month or $2000 a year were included. Within these boundaries, the researchers tried to obtain a random sample. 13. Covering 12,096 ‘‘white’’ families, in 92 cities or localities in 42 states. The aim of the survey was ‘‘to get representative data that would show living conditions in all sections of the country and in all kinds of localities.’’ 14. A survey among 92 families in Toulouse, between 1936 and 1938. It is unclear how representative the sample is, and the numbers seem too small to yield a robust outcome. Nevertheless, since few other sources on family entertainment expenditure exist, it is used to extract information about orders of magnitude and general expenditure patterns for entertainment expenditure. The 92 families all consisted of married couples; single-person households were left out. The families were divided into two social groups, the working class, ‘‘ouvriers’’, and the middle class, ‘‘employees’’, and into four income groups: those with an annual family income below 1,200 francs, those with an income between 1,200 and 1,800 francs, those with a family income above 1,800 francs, with an average income of 2,000 francs and, finally, ‘‘the rich’’, with an average income of 3,700 francs. Because it is unclear how the difference between working class and employees was exactly defined, and how the social difference mattered, as well as to obtain a larger sample, here the social groups are taken together, and income is used as the sole criterion for the four sample groups. 15. During the late nineteenth and twentieth century, Britain experienced a sharp increase in income equality, which may have affected the habits of entertainment consumption, and give some credence to the reasoning above. The Gini coefficient
134
GERBEN BAKKER
decreased from 0.52 in 1867, or between 0.47 and 0.59 in 1880, to 0.34 in 1962–1963 (Brenner, Kaelble, & Thomas, 1991, p. 26). Likewise, the share of the top 20 per cent in total income decreased from 62 per cent in 1867 and 58 per cent in 1880, to about 50 per cent in the 1930s, and 43 per cent in 1963 (Williamson, 1991, p. 58). For comparison, in the US the share of the top 20 per cent fell from 52 per cent in 1935–1936 to a low of 44 per cent in 1960, after which it increased again (ibid.). 16. lim y ¼
Y !1
a bY
1 þ1
With Y is the income and the line a+bY the estimated expenditure line. 17. Part of the difference may be due to potentially biased samples for Britain and France; national total consumer expenditure estimates, below, show a far smaller difference between Britain and the US, although not between France and the US. 18. For Britain, an estimate of expenditure on and quantity of tickets to sports matches had to be deducted to arrive at comparable data (see the notes to Fig. 3). 19. A scale-independent statement for gamma, which yields the same results in the case above, of course, is: qc g¼ qc þ ql qc q 1 p 1 þ1¼ c ¼ l ql ql g pc g 1 g¼ pc 1 p l
20. Using pl 1 a qc ¼ pc a ql and qc ¼
1 a qc q þ qmax c a ql l
yielding: g¼
1 Pc 1 þ 1a a Pl
21. If we would assume scale effects, a greater preference for live entertainment in a country could lead to lower relative prices. These scale effects appear to be different from the joint effects (of e on pc/pl and vice versa).
The Evolution of Entertainment Consumption
135
22. And in 1990, this would have reached 90 per cent, although this finding is meaningless, as growth rates for the period after 1945 have not been used. 23. It should be emphasized that this is an entirely hypothetical growth/share simulation. If we look at the calendar years at which the quantity consumed drops below 1 per cent of the total quantity of the 5 categories (i.e. leisure goods/services consumed), this would be 1934 for drink, 1947 for printed media, 1967 for sports and finally, 2053 for non-market recreation. 24. At best, without the boom in demand, cinema may have suffered the same fate as the phonograph, which for years remained somewhat of an expensive elite product, both because of its consumers and because of its content (its styles of music). It never reached the same number of consumers as cinema had done. It was only in the 1950s with its affluent teenagers that the phonograph really became a mass product, and the music industry came to look a bit more like the film industry. See, for example, Bakker (2006).
ACKNOWLEDGEMENTS The author would like to thank Marina Bianchi, Michael Haines, Paul Johnson, Jaime Reis, Ulrich Witt and the anonymous referees for comments and suggestions. The paper also strongly benefited from the comments and suggestions of the participants of the workshop ‘Economic Theory and the Practice of Consumption: Evolutionary and other Approaches’, organised by the University of Cassino and the Max Planck Institute, 18–20 March 2005. The author alone, of course, is responsible for remaining errors. Research for this paper was partially supported by an ESRC AIM Ghoshal Research Fellowship, grant number RES-331-25-3012.
REFERENCES Abel, R. (1999). The red rooster scare. Making cinema American, 1900–1910. Berkeley: University of California Press. Allen, R. C. (1979). Motion picture exhibition in Manhattan, 1906–1912: Beyond the Nickelodeon. Cinema Journal, 18(2), 2–15. Allen, R. C. (1980). Vaudeville and film, 1895–1915. New York: Arno Press. Bakker, G. (2001a). Stars and stories: How films became branded products. In: Enterprise and society (Vol. 2, No. 3, pp. 461–502). Bakker, G. (2001b). Entertainment industrialised. The emergence of the international film industry, 1890-1940. Florence: European University Institute, Ph.D. dissertation. Bakker, G. (2003). Building knowledge about the consumer: The emergence of market research in the motion picture industry. In: Business history (Vol. 45, No. 1, pp. 101–127); reprinted in A. Godley & R. Church (Eds), The emergence of modern marketing. London: Frank Cass. Bakker, G. (2004). At the origins of increased productivity growth in services: Productivity, social savings and the consumer surplus of the film industry, 1900–1938. Working Papers in Economic History, London School of Economics, No. 81.
136
GERBEN BAKKER
Bakker, G. (2005a). The decline and fall of the European film industry: Sunk costs, market size and market structure, 1895–1926. Economic History Review, 58(2), 311–352. Bakker, G. (2005b). The economic history of the international film industry. In: R. Whaples (Ed.), Eh.net Encyclopedia. http://eh.net/encyclopedia/article/bakker/film Bakker, G. (2006). The making of a music multinational: PolyGram’s international businesses, 1945–1998. Business History Review, 80(1), 81–123. Bordwell, D., Staiger, J., & Thompson, K. (1985). The classical Hollywood cinema. Film style and mode of production to 1960. New York: Columbia University Press. Brady, D. (1972). Consumption and the style of life. In: L. Davis (Ed.), American economic growth (pp. 61–92). New York: Harper & Row. Brenner, Y. S., Kaelble, H., & Thomas, M. (1991). Income distribution in historical perspective. Cambridge: Cambridge University Press. Burrows, J. (2004). Penny pleasures: Film exhibition in London during the Nickelodeon era, 1906–1914. In: Film History (Vol. 16, No. 1, pp. 16–91). David, P. A., & Solar, P. (1977). A bicentenary contribution to the history of the cost of living in America. In: Research in economic history (Vol. 2, pp. 1–80). Economic club, Family budgets. Being the income and expenditure of 20 British households 18911894 (London, Economic Club, 1896). Elsaesser, T. (Ed.) (1990, 1994). Early cinema: Space, frame, narrative. London: British Film Institute. Feinstein, C. H., & Thomas, M. (2002). Making history count: A primer in quantitative methods for historians. Cambridge: Cambridge University Press. Forest, C. (1995). Les derniers seances. Cent ans d’exploitation des salles de cine´ma. Paris: CNRS-Editions. Friedel, R. (1979). Parkesine and celluloid. The failure and success of the first modern plastic. In: History of technology (Vol. 4, pp. 45–62). Gomery, D. (1992). Shared pleasures. A history of movie presentation in the United States. Madison: University of Wisconsin Press. Haines, M. (1979). Industrial work and the family life cycle, 1889–1890. In: Research in economic history (Vol. 4, pp. 289–356). Hannan, M. T., & Freeman, J. (1989). Organizational ecology. Cambridge, MA: Harvard University Press. Hansen, M. (1991). Babel and Babylon. Spectatorship in American silent film. Cambridge, MA: Harvard University Press. Horrell, S. (1996). Home demand and British industrialization. Journal of Economic History, 3, 561–604. Jowett, G. S. (1974). The first motion picture audiences. Journal of Popular Film, 3, 39–45. Ko¨nig, W., & Weber, W. (1990). Propyla¨n Technikgeschichte. Netzwerke, Stahl und Strom, 1840 bis 1914 (pp. 527–530). Berlin: Propyla¨n Verlag. Koszarski, R. (1990). An evening’s entertainment. The age of the silent feature picture, 1915–1928 (pp. 13–15). New York: Scribner. Le Play, F. (1877), Les ouvriers Europe´ens. E´tudes sur les travaux, la vie domestique et la condition morale des populations ouvrie`res de l’Europe. D’a`pre`s les fait observe´s de 1829 a` 1855. Avec des epilogues indiquant les changements survenus depuis 1855 (Tours, Mame, second edition, 1877). Low, R., & Manvell, R. (1948). The history of the British film, 1896–1906. London: Allen & Unwin. Meusy, J.-J. (1995). Paris-palaces ou le temps du cinema, 1894–1918. Paris: CNRS-Editions.
The Evolution of Entertainment Consumption
137
Meusy, J.-J. (2002). How cinema became an industry: The big boom in France between 1905 and 1908. Film History, 14(4), 418–429. Michaelis, A. R. (1958). The photographic arts. Cinematography. In: C. Singer (Ed.), A history of technology. The late nineteenth century, c. 1850 to c. 1900 (Vol. V, pp. 734–751). Oxford: Clarendon Press, reprint 1980. Mitchell, B. R. (1993). Historical Statistics Europe. London: Macmillan. Mokyr, J. (1990). The lever of riches. Technological creativity and economic progress (pp. 275–278). Oxford: Oxford University Press. Musser, C. (1990). The emergence of cinema. The American screen to 1907 (pp. 17–20). New York: Scribner. Nelson, R. R., & Winter, S. G. (1982). An evolutionary theory of economic change. Cambridge, MA: Belknap Press of Harvard University Press. Nystrom, P. H. (1931). Economic principles of consumption. New York: Ronald Press. Owen, J. D. (1970). The price of leisure. An economic analysis of the demand for leisure time. Montreal: McGill-Queens University Press. Prest, A. R. (1954). Consumer expenditure in the United Kingdom, 1900–1919. Cambridge: Cambridge University Press. Redi, R. (1995). Le cine´ma Italien 1909–1920: L’expansion et le reflux. In: R. Cosandey & F. Albera (Eds), Cine´ma sans frontie`res/Images across borders 1896–1918. Quebec/ Lausanne: Nuit Blanche/Editions Payot. Richards, J. (1994). Cinema going in worktown. Historical Journal of Film Radio and Television, 14(2), 147–166. Ross, S. J. (1998). Working-class Hollywood. Silent film and the shaping of class in America. Princeton: Princeton University Press. Sadoul, G. (1962). Le cine´ma franc- ais, 1890–1962. Paris: Flammarion. Sedgwick, J. (1998). The nature of film as a commodity. Discussion papers in business economics, University of North London, No. 20. Sedgwick, J. (2000). Popular film in 1930s Britain. A choice of pleasures. Exeter: University Press of Exeter Press. Singer, B. (1995). Manhattan Nickelodeons: New data on audiences and exhibitors. Cinema Journal, 34(3), 5–35. Sklar, R. (1975). Movie-made America. A cultural history of American movies. New York: Vintage Books (revised edition, 1993). Staiger, J. (1990). Announcing wares, winning patrons, voicing ideals. Thinking about the history and theory of film advertising. Cinema Journal, 29(3), 3–31. Stone, R. (Ed.) (1966). The measurement of consumer expenditure and behaviour in the United Kingdom, 1920–1938, Volume II. London: National Institute of Economic and Social Research. Taylor, A. J. P. (1976). English history, 1914–1945 (revised. edition). London: Oxford University Press. United States Department of Commerce (1975). Historical statistics of the United States from colonial times to 1970. Washington: United States Department of Commerce. Williamson, J. G. (1991). British inequality during the industrial revolution. accounting for the Kuznets Curve. In: Y.S. Brenner, H. Kaelble & M. Thomas (Eds), Income distribution in historical perspective (pp. 57–75). Cambridge: Cambridge University Press. Zimmerman, C. C. (1936). Consumption and standards of living. New York: D. Van Nostrand Company Inc.
This page intentionally left blank
CINEMA AND TV: AN EMPIRICAL INVESTIGATION OF ITALIAN CONSUMERS$ Andrea Sisto and Roberto Zanola ABSTRACT A number of papers have empirically investigated the demand for cinema by applying the rational addiction model proposed by Becker and Murphy (1988). However, they fail to take account of the relationship between movie and television consumption. The purpose of this paper is to extend previous works on cinema demand by including both cinema and television movie consumption. To this aim a panel-data generalized method of moments (GMM) methodology is used to estimate a dynamic model of double rational addiction as proposed by Bask and Melkersson (2004) using a sample of monthly time- and cross-sectional series covering the 20 Italian regions over the period 2000–2002.
$
An earlier version of this paper was presented at Workshop on the Evolution of Consumption, Gaeta (Italy). This paper has benefited from comments by the participants, and by helpful discussions with M. Bianchi, A.E. Scorcu, B. Frey, and J. Sedgwick. The usual disclaimers apply.
The Evolution of Consumption: Theories and Practices Advances in Austrian Economics, Volume 10, 139–154 Copyright r 2007 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 1529-2134/doi:10.1016/S1529-2134(07)10006-5
139
140
ANDREA SISTO AND ROBERTO ZANOLA
1. INTRODUCTION Dynamics have been shown to be very important for the explanation of individual behaviour in many empirical specifications. Hence, it seems natural to model the demand for addictive goods using a model which allows linkages of consumption over time. Becker and Murphy’s (1988) rational addiction model explains the consumption of addictive goods as a specific dynamic human capital accumulation process. Whatever the origin of addiction (pharmacological or psychological), the framework used to deal with habit formation (endogeneity of preferences or shadow prices) and the assumption about consumer behaviour (rationality or myopia), the key feature of this model is the deterministic impact of past consumption on current behaviour. Although this model has been empirically investigated by a number of studies (McCain, 1979, 1995; Villani, 1992; Abbe´-Decarroux, 1994; Le´vy-Garboua & Montmarquette, 1996), only few authors have analysed positive rational addiction, such as the case of cinema demand. Cameron (1999) and Dewenter and Westermann (2005) apply the Becker and Murphy’s (1988) rational addiction model to a single country, respectively UK and Germany, failing to provide support for all the predictions of the rational addiction models. Sisto and Zanola (2004) extend previous works by analysing a panel of thirteen European countries. Results provide strong evidence in favour of a model of cinema demand that emphasizes the role of past and future consumption in current consumption. Despite these findings, however, new useful insights might be provided through an investigation of the relationship between consumption of cinema and television. First, if cinema and television are important substitutes (or complements), then a correctly specified cinema demand equation must include television. Otherwise, the estimated coefficients of the included variables may be biased, depending, as usual, on the relationship between the included and excluded variables. Secondly, models that include multiple addictive goods may offer useful policy guidance to evaluate the impact of public interventions since it is not sufficient to consider the consumption of different addictive goods independently (Decker & Schwarts, 2000; Palacios-Huerta, 2003). The purpose of this paper is to extend previous works on cinema demand by including two addictive consumption goods, cinema and television movie consumption. The idea of specific human capital accumulation in movie consumption underlying the rational addiction approach is extended by allowing television consumption to reinforce the movie stock consumption
141
Cinema and TV
(in the case of a complementary effect) or to depreciate it (if substitution effect emerges). To this aim a panel-data generalized method of moments (GMM) methodology is used to estimate a dynamic model of double rational addiction as proposed by Bask and Melkersson (2004) using a sample of monthly time- and cross-sectional series covering the 20 Italian regions over the period 2000–2002. The proceeding of this paper is as follows. Section 2 presents the theoretical framework. Data are described in Section 3. The econometric methodology is shown in Section 4. Section 5 summarizes the empirical findings. Section 6 concludes.
2. THEORETICAL MODEL Following Becker and Murphy’s rational addiction theory, a consumer is said to be addicted if an increase in past consumption leads to an increase in current consumption ceteris paribus. This type of behaviour involves reinforcement, as an increase in past level of consumption increases the marginal utility for present consumption, and tolerance, as the satisfaction from a given level of consumption is lower when past level is greater (Bask & Melkersson, 2004). Although tolerance appears to be a reasonable assumption in the case of harmful goods (alcohol, tobacco, drugs, and so on), as a rational addicted consumer would discount the harmful effect of future addiction, it is more difficult to understand how it works in the case of harmless goods, such as the case of cinema consumption. People, in fact, can be addicted not only to harmful goods, but also to activities that may seen to be physically harmless. When this occurs, the current consumption of the good increases the future consumption of the good, and this occurs ‘‘only when the past consumption of the good raises the marginal utility of present consumption’’ (Becker & Murphy, 1988). If applied to cinema consumption the argument could be made that people will become addicted to cinema if their past participation has led to an accumulation of cinema capital which allows them to raise the marginal utility of present participation. This addiction will in turn lead to increased participation in future. However, unlike negatively addictive goods, beneficially addictive goods have a high elasticity of demand, and are thus very sensitive to price. Hence, the question here is whether cinema prices have changed so much as to justify ‘tolerance’. If we limit our analysis to ticket prices, the answer to previous question is necessarily negative. However,
142
ANDREA SISTO AND ROBERTO ZANOLA
cinema is a time-consuming activity. Hence, the optimal path of cinema consumption might increase at diminishing rates as opportunity costs of going to cinema are expected to increase with age, due to the (opportunity) cost of time and the access to less time-consuming alternatives, such as television consumption. In fact, television viewing provides the possibility for different activities beside it and, furthermore, it offers some competitive advantages like comfort, convenience, privacy together with a potentially higher variety and quality of supplied programs (Cameron, 1990; Ferna´ndez Blanco & Ban˜os Pino, 1997; Dewenter & Westermann, 2005). The Becker and Murphy’s model of addiction considers only one addictive good and one non-addictive good. In what follows we extend this basic model by including two addictive goods, rather than one addictive, along with one non-addictive good (Bask & Melkersson, 2004). Relaxing separability by allowing utility in each period to depend on consumption in the current and previous periods (Becker, Grossman, & Murphy, 1994), a concave utility function at time t may be defined as U t ¼ U ðC t ; C t1 ; At ; At1 ; Y t ; et Þ
(1)
where Ct and At are respectively the consumption of two addictive goods at time t, Yt the consumption of non-addictive good, and et represents all the other factors affecting individual’s utility. Assuming a constant rate of time preference, s, a constant price of the non-addictive good, treated as numeraire, and perfect capital markets, the intertemporal decision of the individuals consists in maximizing utility subject to a usual intertemporal budget constraint as follows:
U¼
T X
st1 U ðC t ; C t1 ; At ; At1 ; Y t ; et Þ
t¼0
s:t:
T X
rt1 ðY t þ Pct C t þ Pat At Þ W t
ð2Þ
t¼1
where r is the constant real interest rate, Pc and Pa the prices of the two addictive goods, and Wt the value of earnings at time t. A standard technique used in literature to derive demand equations is to approximate the instantaneous utility function in the neighbourhood of
143
Cinema and TV
steady state by a quadratic function in the arguments. Assuming a rate of time preference equal to the market interest rate, one can derive the following difference equation: C t ¼ y1 C t1 þ y2 C tþ1 þ y3 Pct þ y4 At1 þ y5 At þ y6 Atþ1 þ y7 Pat þ y8 et þ y9 etþ1
ð3Þ
where the ys are parameters which depend on the underlying preferences. Eq. (3) nests different behaviours (Bask & Melkersson, 2004). Since a good is addictive if and only if an increase in past consumption leads to an increase in current consumption, a zero value of both y1 and y4 implies a non-addicted consumer. Testing for rational addiction is testing for forward-looking behaviour. An addicted but myopic consumer is not farsighted, according to a simply backward looking consumer decision, while a rational addicted consumer takes into account consequences of past, current, and future information. The rational model implies that coefficients on future consumption should have the same sign as coefficients on lagged consumption (the sign only differs by the term d).
3. DATA This section provides a brief description of the data sets used in this paper. The data consist of aggregate regional monthly time series from January 2000 to December 2002 for the 20 Italian regions. In particular, the theoretical framework is investigated using the following variables. (i) Cinema demand: cinema demand, adm, is captured by the number of tickets sold in one month divided by population size. Data are obtained from European Cinema Yearbook published by Mediasalles and Lo spettacolo in Italia published by SIAE. (ii) Television consumption: the identification of an appropriate measure of television consumption is a difficult task. Previous empirical researches use very different strategies to deal with this issue. Cameron (1986) uses the number of colour licences; Cameron (1990) defines a log dummy time trend to account for private television channels introduction; Ferna´ndez Blanco and Ban˜os Pino (1997) employ a dummy variable to model the introduction of regional television channels in Spain; Macmillan and Smith (2001) adopt the number of licences for
144
ANDREA SISTO AND ROBERTO ZANOLA
households to capture the effect of TV-set introduction on post-war cinema admissions. However, all these strategies appear to be imperfect measures of TV consumption and, consequently, empirical analyses are likely to be biased. In fact, inclusion of dummy variables could not fully capture either changes in TV ownership nor in quality/variety of television programming through the availability of more private or regional channels. The use of television licenses is also problematic. First, licenses are paid only on the first TV-set owned so that this measure substantially underestimate the number of total household TV-sets existing. Secondly, it does not provide any valid measure of consumption. A way to overcome some of these limits is to use TV audience. Hence, in order to check for a common habit accumulation process between cinema and TV, we use public and commercial television1 prime time movie audience, TV, calculated as the monthly number of TV movie spectators divided by population size. Data are obtained from Auditel data elaboration published by Media Consultants. (iii) Price: price of cinema admission, p, is the ticket price at box-office.2 The variable is the average expense per film-goer, which is the ratio between the monthly total regional receipts and the monthly number of film-goers, according to the data supplied by Mediasallers and SIAE publications. One would expect that the cinema demand is negatively related to the ticket price. However, in our case the elasticity of cinema demand to price may not be overinterpreted due to the difficulty to exactly capture price variations when using monthly data (Dewenter & Westermann, 2005). By contrast, a pecuniary measure of TV price is difficult to define. In fact, Italian television owners have a virtually zero marginal costs for movie consumption as the license cost must be incurred whether or not the television set is used. Hence, the effect of TV movie price is compressed in the error term. (iv) Other factors: disposable per capita income and number of screens are other important factors affecting cinema demand (Macmillan & Smith, 2001; Sisto & Zanola, 2004). However, since monthly observations prevent us from taking account of these variables, the effects of them on cinema consumption is partially captured by regional dummies.
All monetary variables are deflated at 1995 price level by CPI. Descriptive statistics are reported in Table 1.
145
Cinema and TV
Table 1. Variable adm
p
TV
Definition of the Variable Used in the Study.
Definition and Source
Mean
Standard Deviation
Minimum
Maximum
Regional monthly cinema ticket sold per capita S.I.A.E., Lo spettacolo in Italia (2001); Mediasalles, European Cinema Yearbook (2001–2003); Istat, Statistiche demografiche (2000– 2002) Regional monthly average cinema ticket price at 1995 price levels S.I.A.E., Lo spettacolo in Italia (2001); Mediasalles, European Cinema Yearbook (2001–2003); Media consultants, AUDITEL data elaboration Regional average prime time movies audience
0.146
0.094
0.011
0.532
4.558
0.600
2.952
6.728
0.084
0.062
0.062
0.310
4. EMPIRICAL METHODOLOGY In order to apply the theoretical framework of previous section, different specifications of Eq. (3) are estimated: admi;t ¼ a0 þ a1 admi;t1 þ a2 admi;tþ1 þ a3 pi;t þ i;t admi;t ¼ b0 þ b1 admi;t1 þ b2 admi;tþ1 þ b3 pi;t þ b4 TV i;t þ i;t admi;t ¼ l0 þ l1 admi;t1 þ l2 admi;tþ1 þ l3 pi;t þ l4 TV t1 þ l5 TV i;t þ l6 TV i;tþ1 þ i;t
ð4Þ
where subscripts i and t stand respectively for the region and the period (month) considered and ei,t is the error term. The direct estimation of Eq. (4) may give rise to some misleading results and some caution is necessary. First step in the analysis is to test whether
146
ANDREA SISTO AND ROBERTO ZANOLA
data can be pooled. Following Levaggi and Zanola (2003), F-tests are performed on the null hypothesis that the coefficients for each variable in Eq. (40 ) are the same for each year. Results are reported in Table 2. The null hypothesis of equal coefficients could not be rejected in either case, therefore data can be pooled. Next step is to check for stationarity of the variables included in the model. In order to determine the presence of panel unit root, we apply a battery of tests running the IPS t-bar tests (Im, Pesaran, & Shin, 2003) and the Levin–LinChu test (Levin, Liu, & Shin, 2002), including a heterogeneous time trend in each specification. Although results in Table 3 seem to strictly reject the null hypothesis of nonstationary for all the variables of the model, in some cases tests have
Table 2.
Testing Pooling Restrictions.
Variable Test for pooling DY*pi,t DY*TVi,t DY
F-test
P-value
0.77 0.48 0.22
0.823 0.995 1.000
Notes: for testing the hypothesis of pooling the following augmented models have been estimated: X admi;t ¼ b0 þ b1 Padmi;t þ b1i DY Padmi;t X X DY þ i;t ð40 Þ þ b2 TV i;t þ b2i DY TV i;t þ b3
by adding T1 monthly dummy for each variable that take the value 1 if observation belongs to the month considered. The F-test is performed on the coefficient of these variables.
Table 3. Unit Root Tests for Panel. Variable Im–Pesaran–Shin test (2003) adm p TV Levin–Lin–Chu test (2002) adm p TV
Trend
t-bar
p-value
2.463 2.819 3.416
0.071 0.001 0.000
Yes Yes Yes
0.3699 0.3492 0.6515
0.024 0.000 0.000
Yes Yes Yes
147
Cinema and TV
proved to be sensitive to the number of lags and to time trend. Banerjiee (1999) showed that both IPS and Levin–Lin–Chu tests suffered from low power and failed to distinguish between I(0) and I(1). Visual inspection of each variable suggests a strong presence of seasonality, inducing us to include monthly dummy variables to prevent our estimates from spurious regressions. Therefore, Eq. (4) is estimated in first differences, introducing 11 monthly dummies to check for seasonality, as follows: Dadmi;t ¼ a0 þ a1 Dadmi;t1 þ a2 Dadmi;tþ1 þ a3 Dpi;t þ i;t Dadmi;t ¼ b0 þ b1 Dadmi;t1 þ b2 Dadmi;tþ1 þ b3 Dpi;t þ b4 DTV i;t þ i;t Dadmi;t ¼ l0 þ l1 Dadmi;t1 þ l2 Dadmi;tþ1 þ l3 Dpi;t þ l4 DTV t1 þ l5 DTV i;t þ l6 DTV i;tþ1 þ i;t
ð5Þ
where D is the first difference operator.
5. RESULTS Taking first differences will induce a first-order moving average process into the transformed residuals (Arellano, 1989). Hence, in order to get consistent estimates we instrument endogenous variables (past and future consumption of dependent variable) with two and three lagged and lead levels of ticket price variable. Owing to over-identification, we adopt the GMM which is proved to be an appropriate method to get a consistent estimator when the number of instruments is higher than exogenous variables, as is the case here. The GMM estimator has the further advantage that it does not need the restrictive assumption of strictly exogeneity of the independent variables since they may be assumed to be predetermined or endogenous (Heinesen, 2004). Instruments validity is checked using the Sargan test for overidentifying restrictions (Sargan, 1958; Hansen, 1982). Table 4 summarizes the main results. Column 2 reports the results of the standard Becker and Murphy’s (1988) model of addiction which considers only one addictive good and one nonaddictive good. Results provide a strong evidence that cinema consumption conforms to a rational addiction hypothesis. Both the coefficients on past and future consumption are positive and significantly different from zero, so that past and future changes significantly impact on current consumption. Furthermore, we also notice that a1>a2 as expected. This finding may support the hypothesis that cinema is a time consuming activity. In fact, the existence of increasing opportunity cost related to age, together with the
148
ANDREA SISTO AND ROBERTO ZANOLA
Table 4.
Cinema Demand Estimates.
Variable
GMMa
GMMb
adm(t1)
0.4604 (0.0544) 0.2524 (0.0353) –
0.4702 (0.0586) 0.2548 (0.0411) –
TV(t)
–
TV(t+1)
–
0.1862 (0.0702) –
0.0878 (0.0292) 0.0011 (0.0002) 30.653 (0.05995) 0.9098 – – 0.3728 0.79621
0.0989 (0.0283) 0.0378 (0.0208) 26.930 (10.628) 0.9110 – – 0.4250 0.9367
adm(t+1) TV(t-1)
P Const Hansen R2 C test for TV (t-1) C test for TV (t+1) Zs Zl
GMMc
GMMd
0.4488 (0.0631) 0.2413 (0.0455) 0.0359 (0.0490) 0.2023 (0.0736) 0.1185 (0.0631) 0.1008 (0.0269) 0.0097 (123.94) 26.353 (0.12066) 0.9124 2.807 0.851 0.4133 0.8471
0.4331 (0.0624) 0.2206 (0.0461) 0.3459 (0.1759) 0.2935 (0.0895) 0.2864 (0.1165) 0.1365 (0.0333) 0.0665 (0.0246) 22.796 (0.24644) 0.9021 – – 0.5287 1.0266
a
Becker and Murphy rational addiction model. Inclusion of current TV audience. c Inclusion of both past and future TV audience treated as exogenous regressors. d Only past TV audience is treated as endogenous. Significance coefficient at 0.01. Significance coefficient at 0.05. Significance coefficient at 0.10. b
access to less time consuming leisure activities, decreases the impact of future consumption, so that the coefficient is positive, but smaller than that associated to past consumption. Also the price coefficient is negative and significantly different from zero. The short-run price elasticity3 is smaller than that obtained in comparable studies of rational addictive products (Dewenter, 2003). The long-run price elasticity is higher than the short-run one, as predicted in Becker et al. (1994). The second specification includes present TV movie consumption. Column 3 shows the results of this specification. Here too, both past and future cinema consumption seems to affect present consumption. Thus, this outcome confirms the hypothesis of rational behaviour. This specification also
Cinema and TV
149
provides evidence on the relationship between cinema consumption and TV movie consumption. The effect of TV movie consumption is positive and statistically significant. This is taken as an evidence of a complementary relationship between cinema and TV consumption. The short- and the longrun elasticities are similar to the previous specification, respectively 0.42 and 0.94. Column 4 shows the results of the most general specification, which includes both past and future TV movie consumption as explanatory variables (Bask & Melkersson, 2004). To check for possible endogeneity of TV movie consumption a C-test has been computed for both past and future TV movie consumption. Findings suggest the endogeneity of past TV variable. Hence, in column 5 only past TV movie audience is treated as endogenous. Again, past and future cinema consumption are significant and positively related to present cinema consumption. Moreover, findings are still consistent with the evidence that cinema and TV movie consumption are complements and the short- and the long-run elasticities display values similar to other studies (Ferna´ndez Blanco & Ban˜os Pino, 1997; Dewenter & Westermann, 2005). Although results from Table 4 confirm the presence of a common habit formation process, however it seems interesting to discriminate between weekend and weekday TV movie consumption. Table 5 displays the results of the dynamic regression between cinema admission and weekend TV prime-time movie consumption, week.4 Three different specifications are estimated. Column 2 shows the results of the rational addiction model with the inclusion of current TV audience. The model which includes both past and future TV audience treated as exogenous variables is presented in column 4. Finally, performed C-test suggests the endogeneity of past TV variable so that only future TV movie audience is treated as endogenous in column 4. Estimates show evidence in favour of a substitution relationship during weekend. Both lag and lead value of TV movie consumption exhibit a negative impact on current cinema demand while the coefficient associated with present TV movie consumption does not appear statistically significant at all common levels. Results confirm the presence of rational addiction in cinema consumption. As usual, long-run price elasticity exceeds short-run price elasticity. Finally, Table 6 considers the relationship between cinema demand and weekday TV prime-time movie consumption, daily.5 Coefficients seem to be quite similar to those described in Table 4. In fact, an increase in weekday TV movie consumption seems to stimulate a higher appetite for cinema consumption; these results confirm the presence of a common habit stock accumulation.
150
Table 5.
ANDREA SISTO AND ROBERTO ZANOLA
Cinema Demand and Weekend Prime-Time Movie Audience.
Variable
GMMa
GMMb
GMMc
adm(t1)
0.5015 (0.0713) 0.2408 (0.0489) –
0.5392 (0.0757) 0.2655 (0.0503) 0.0413 (0.0515) 0.2219 (0.0656) 0.0694 (0.0358) 0.1390 (0.0294) 0.0670 (0.0213) 22.175 (0.3311) 0.9092 1.632 6.337 0.6123 1.6334
0.6119 (0.0917) 0.2586 (0.0638) 0.1550 (0.0740) 0.1256 (0.0809) 0.2870 (0.1150) 0.1550 (0.0317) 0.0788 (0.0229) 13.221 (0.8271) 0.8973 – – 0.7042 2.1960
adm(t+1) week(t1) week(t) week(t+1) padm const Hansen J stat2 R2 C test for week(t1) C test for week(t+1) Zs Zl
0.2338 (0.0607) – 0.1342 (0.0289) 0.0045 (70.5094) 25.173 (0.1551) 0.9119 – – 0.5364 1.2880
a
Becker and Murphy rational addiction model with the inclusion of current TV audience. Inclusion of both past and future TV audience treated as exogenous variables. c Only future TV audience is treated as endogenous. Significance coefficient at 0.01. Significance coefficient at 0.05. Significance coefficient at 0.10. b
In summary, estimates provide evidence for a complementary relationship between cinema and TV movie consumption, with the exception of weekend TV movie consumption. Although at first sight these results seem to be in contrast with empirical findings of economic literature, since television has always been seen as one of the key determinants in cinema demand fall across Europe during seventy, however there is a plausible explanation for this. In fact, it is likely that TV movie consumption might be simultaneously complementary since the appetite for cinema consumption is going to grow with TV movie consumption, but also substitute to cinema in terms of time allocation. Again, the existence of increasing opportunity cost related to age, together with the access to less time consuming leisure activities, decreases the impact of future consumption.
151
Cinema and TV
Table 6. Cinema Demand and Weekday Prime-time Movie Audience. Variable
GMMa
GMMb
GMMc
adm(t1)
0.4670 (0.0545) 0.2589 (0.0370) –
0.4427 (0.0653) 0.2480 (0.0458) 0.0333 (0.0318) 0.1054 (0.0571) 0.2003 (0.0667) 0.1025 (0.0268) 0.0401 (0.0198) 25.572 (0.1805) 0.9120 5.046 0.597 0.4045 0.8192
0.4434 (0.0642) 0.2174 (0.0476) 0.3111 (0.1295) 0.2333 (0.0814) 0.2914 (0.0797) 0.1444 (0.0329) 0.0716 (0.0245) 20.151 (0.3855) 0.8971 – – 0.5294 1.0688
adm(t+1) daily(t1) daily(t) daily(t+1) padm const Hansen J stat3 R2 C test for daily(t1) C test for daily(t+1) Zs Zl
0.0785 (0.0528) – 0.0911 (0.0289) 0.0287 (0.0210) 29.203 (0.0838) 0.9101 – – 0.3752 0.8219
a
Becker and Murphy rational addiction model with the inclusion of current TV audience. Inclusion of both past and future TV audience treated as exogenous variables. c Only past TV audience is treated as endogenous. Significance coefficient at 0.01. Significance coefficient at 0.05. ***Significance coefficient at 0.10. b
These findings seem to suggest a widely interpretation of the rational addiction model. In fact, in Becker and Murphy’s model of addiction the stock component of the shadow, or full, price is itself endogenously determined by the optimal path, and yet it can be said to help determine the optimal path by affecting the cost of the addictive good. However, the concept of price needs to be handled very carefully. The full of addictive goods to rational consumers includes the money value of changes in future utility and earnings induced by changes in current consumption. In our model, by contrast, the interpretation of the relative price of time on habit formation as a component of the full price, allow us to use the model to analyse a widely range of goods. However, a word of caution must be used in interpreting these results. In fact, the strength of this finding is
152
ANDREA SISTO AND ROBERTO ZANOLA
somewhat weakened by the procedure used to deal with ticket prices. Since data are aggregated, prices are built as yearly average admission costs which prevent us from capturing price variation across different days and categories of purchasers. Analogously, the typical seasonal trend registered in box office revenue cannot be captured using annual data.
6. CONCLUSION Interest in the rational addiction model has expanded to some goods where no pharmaceutical dependency arises, such as the case of cinema. The purpose of this paper was to empirically investigate the demand for cinema by focusing on the role of TV audience on cinema consumption. To this aim a panel-data GMM methodology is used to estimate a dynamic model of double rational addiction as proposed by Bask and Melkersson (2004) using a sample of monthly time- and cross-sectional series covering the 20 Italian regions over the period 2000–2002. Results provide strong evidence in favour of a model of cinema demand that emphasizes the role of past and future consumption on current consumption. Furthermore, estimates provide evidence for a complementary relationship between cinema and TV, even if cinema and TV movie consumption seem to be substitute when weekend and weekday TV movie consumption are distinguished. However, further investigation is still required. In particular, a promising direction for future research might be to explore individual rather than aggregated data in order to capture price variation across different days and categories of purchasers. To this aim, a promising field of research seems to analyse the demand for DVD.
NOTES 1. RAI and the two major commercial broadcasting networks (Mediaset and La Sette). 2. Owing to the lack of information, we cannot take account of a second component of price – the cost of other activities which are gradually becoming essential to cinema attendance – for a description of which see Ferna´ndez Blanco and Ban˜os Pino (1997). 3. The short- and long-run elasticities, respectively Zs and Z1, are computed as suggested in Dewenter (2003). 4. Weekend TV prime-time audience is measured as monthly holiday prime-time movie audience (Saturday, Sunday, and international holidays). 5. Daily audience is measured as average daily TV movie audience excluding weekend and international holidays.
153
Cinema and TV
REFERENCES Abbe´-Decarroux, F. (1994). The perception of quality and demand for services. Empirical application to the performing arts. Journal of Economic Behavior and Organization, 23(1), 99–107. Arellano, M. (1989). A note on the Anderson-Hsiao estimator for panel data. Economics Letters, 31(4), 337–341. Banerjiee, A. (1999). Panel data unit roots and cointegration: An overview. Oxford Bulletin of Economics and Statistics, 61, 607–629. Bask, M., & Melkersson, M. (2004). Rationally addicted to drinking and smoking?. Applied Economics, 36(4), 373–381. Becker, G. S., Grossman, M., & Murphy, K. M. (1994). An empirical analysis of cigarette addiction. American Economic Review, 84(3), 396–418. Becker, G. S., & Murphy, K. M. (1988). A theory of rational addiction. Journal of Political Economy, 96(4), 675–700. Cameron, S. (1986). The supply and demand for cinema tickets: Some UK evidence. Journal of Cultural Economics, 10(1), 38–62. Cameron, S. (1990). The demand for cinema in United Kingdom. Journal of Cultural Economics, 14(1), 35–47. Cameron, S. (1999). Rational addiction and the demand for cinema. Applied Economic Letters, 6(9), 617–620. Decker, S. L., & Schwarts, A.E. (2000). Cigarettes and alcohol: Substitutes or complements?, NBER WP 7535. Dewenter, R. (2003). Rational addiction to news? Habit formation and print media usage, University FAF, Economics Discussion Paper 2. Dewenter, R., & Westermann, M. (2005). Cinema demand in Germany. Journal of Cultural Economics, 3, 213–231. Ferna´ndez Blanco, V., & Ban˜os Pino, J. (1997). Cinema demand in Spain: A cointegration analysis. Journal of Cultural Economics, 21, 57–75. Hansen, L. P. (1982). Large sample properties of generalized method of moments estimation. Econometrica, 50, 1029–1054. Heinesen, E. (2004). Determinants of local public school expenditure: A dynamic panel data model. Regional Science and Urban Economics, 34, 429–453. Im, K. S., Pesaran, M. H., & Shin, Y. (2003). Testing for unit roots in heterogeneous panels. Journal of Econometrics, 115, 53–74. Levaggi, R., & Zanola, R. (2003). Flypaper effect and sluggishness: Evidence from regional health expenditure in Italy. International Tax and Public Finance, 10, 535–547. Levin, A., Liu, C., & Shin, Y. (2002). Testing for unit roots in heterogeneous panels. Journal of Econometrics, 115(1), 53–74. Le´vy-Garboua, L., & Montmarquette, C. (1996). A microeconomic study of theatre demand. Journal of Cultural Economics, 20, 25–50. Macmillan, P., & Smith, I. (2001). Explaining post war cinema admission. Journal of Cultural Economics, 25(2), 91–108. McCain, R. A. (1979). Reflection on cultivation of taste. Journal of Cultural Economics, 3, 30–52. McCain, R. A. (1995). Cultivation of taste and bounded rationality: Some computer simulations. Journal of Cultural Economics, 19, 1–15.
154
ANDREA SISTO AND ROBERTO ZANOLA
Palacios-Huerta, I. (2003). Multiple addiction. Mimeo, Brown University. Sargan, J. D. (1958). The estimation of economic relationships using instrumental variables. Econometrica, 26, 393–415. Sisto, A., & Zanola, R. (2004). Rational addiction to cinema? A dynamic panel analysis of European countries, WP 41, University of Eastern Piedmont. Villani, A. (1992). Rational addiction in the arts. Ricerche Economiche, 46, 41–54.
PART C: SOCIAL COMPETITION AND INTERDEPENDENT PREFERENCES
This page intentionally left blank
SMOKE SIGNALS: ADOLESCENT SMOKING AND SCHOOL CONTINUATION Philip J. Cook and Rebecca Hutchinson INTRODUCTION Smoking initiation by adolescents has been analyzed by economists as a choice reflecting prices, tastes, and subjective evaluation of the long-term risks of addiction and disease. What is missing from this account is the fact that smoking is a social activity and is subject to peer influence. Peers may serve as a source of information about why and how to smoke, and how to obtain cigarettes. Peers also serve as an audience, observing and evaluating others’ behavior. This evaluation is mediated by the long association in popular culture between smoking and a variety of attributes prized by adolescents. Like choice of fashion in hair and clothing, body piercing, comportment, and so forth, smoking by adolescents connotes information about identity. Knowing this, the decision of whether to smoke is partly a decision of what identity to project. Smoking has long been linked in popular culture to such attributes as autonomy, rejection of mainstream values, sophistication, and being ‘‘cool.’’ These cultural links are no doubt strengthened for adolescents by the fact that smoking is illegal for those under 18, banned by many schools, frowned on by adults, and known to be potentially harmful. While these attributes The Evolution of Consumption: Theories and Practices Advances in Austrian Economics, Volume 10, 157–186 Copyright r 2007 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 1529-2134/doi:10.1016/S1529-2134(07)10007-7
157
158
PHILIP J. COOK AND REBECCA HUTCHINSON
make smoking more attractive to many adolescents, it is also true that smoking sends a signal of being off track in school that would generally be perceived by peers as negative. In particular, in this paper we demonstrate that in the United States as of the late 1990s, smoking in 11th grade was a uniquely powerful predictor of whether the student finished high school, and if so whether the student matriculated in a four-year college. For economists, the obvious explanation for this empirical link is interpersonal differences in time preference. Smoking offers immediate benefits (to those who enjoy it) with long delayed costs in the form of disease, disability, and early death. Dropping out of school also offers immediate benefits (for those who would rather have more leisure or earnings) with delayed costs in terms of limited career options. That smokers drop out early is then simply a reflection of their high rate of time discount. This explanation is satisfying because it is based on well-established concepts relating to time preference (Hoppe, 2001; Menger, [1871] 1976; Mises, [1949] 1998; Bo¨hm-Bawerk, [1884–1921] 1959; Strigl, 2001; Rothbard, [2004] 1963). Time preference rates are deemed subjective and to differ among individuals (Smith, 1988, p. 5). These differences have been identified as central to explaining the socioeconomic class structure (Banfield, 1974) and explaining differences in criminal propensities (Banfield, 1974; Banfield, 1977; Wilson & Herrnstein, 1985). What confuses this simple account is our finding that a parallel activity, drinking, is not predictive of school dropout – despite the fact that its temporal payoff is quite similar to smoking. If adolescent decisions about health-related behaviors are generally guided by concern for the future, then why doesn’t adolescent drinking follow the same pattern as smoking? This anomalous finding opens the door to other kinds of explanations, including social status concerns. (While economists may be uncomfortable with this line of explanation, social psychologists would accept status sensitivity as a fundamental of human nature.) When it comes to social status, both the costs and benefits of smoking are immediate but differ among individuals according to whether they would otherwise be perceived by peers as being on track to college, and whether they are in fact on track. We suggest that the decision to smoke is being driven in part by the fact that youths are aware of the messages it conveys to their peers. The smoking signal conveys more information for high-aptitude youths than those with low aptitude, and hence carries a greater social cost for the former. It turns out that both aptitude and commitment to school appear to influence smoking decisions. In what follows, we describe the data set, present results demonstrating the predictive power of smoking and drinking, and then estimate a demand
Smoke Signals: Adolescent Smoking and School Continuation
159
function for smoking by high school juniors that includes whether they are destined to graduate and go on to college. We offer an explanation for the empirical findings that rest on smoking as one signal of being unlikely to continue in school, the power of which is conditioned on aptitude. There are no direct implications from this analysis for whether smoking is in some sense a cause of school dropout. We offer some speculations on this matter in the conclusion.
THE DATA Our data source is the 1997 National Longitudinal Survey of Youth (NLSY), sponsored by the Bureau of Labor Statistics of the US Department of Labor. The sample of 8,984 individuals surveyed is representative of the US population in 1997 that was born between 1980 and 1984 (Center for Human Resource Research, 2002). Data were gathered through hour-long interviews with the participants and questionnaires completed by the parents in that year. Additional data have been collected from the participants every year since; the Labor Department has released these data from the six waves through 2002. The NLSY data include detailed information on respondents’ demographic characteristics, family, work, criminal activities, and health status. Of particular relevance to our work are the items on drinking, smoking, and schooling. Our estimates are based on a sub-sample of 3,915 NLSY respondents who were juniors in high school in 1997, 1998, or 1999. (We excluded those who were relatively old for this grade, and in particular those who were over 18.) Key outcome measures include whether the respondent graduated from high school the year following junior year, and whether the respondent matriculated in a four-year college or university thereafter. Definitions and summary statistics for all variables used in the analysis are presented in Table A1.
SMOKING AS A PREDICTOR OF SCHOOL CONTINUATION Respondents who said that they had smoked at least one cigarette in the previous 30 days were designated ‘‘smokers.’’ Those who had at least one drink of alcohol in the previous 30 days were designated ‘‘drinkers,’’ and
160
PHILIP J. COOK AND REBECCA HUTCHINSON
those who admitted to having five or more drinks on at least four occasions in that period were designated ‘‘frequent binge drinkers.’’ Another key variable in our analysis is the respondent’s score on the math component of the Armed Services Vocational Aptitude Battery (ASVAB). This test was given to most of the NLSY respondents in 1997. Together with modules testing other aptitudes, it was developed by the US military to guide placement of new recruits into alternative specialties. It is also valid as a predictor of school achievement (Roberts et al., 2000). Tables 1 and 2 compare smokers and nonsmokers for various subgroups of respondents with respect to high school graduation rates, and college matriculation rates. For the entire sample, the high school on-time graduation rate for males is 60% for smokers and 75% for nonsmokers; the college matriculation rate is 12% for smokers and 30% for nonsmokers, a more than two-to-one difference. For females, the high school graduation rate is 71% for smokers and 80% for nonsmokers; the college matriculation rates are 21% and 37%.1 The results presented in these tables make clear that the ‘‘smoking signal’’ applies not only to the population of high school juniors as a whole, but also to each of several subgroups. Nonsmokers are more likely than smokers to continue their education: for respondents with low ASVAB as well as those with high; for youths with mothers or fathers who are high school dropouts, as well as those who are college graduates; and for each of three groups defined by race (black, Hispanic, and all others). These results are depicted in Figs. 1 and 2 as well. We refined this analysis by running multivariate logit regressions for males and females. These regressions analyze the effect of smoking on the odds of high school graduation and college matriculation controlling for demographic and family circumstances. Full results are reported in Table A2. Partial results of two specifications are reported in Table 3: the first specification includes the smoker dummy variable, but nothing on drinking, while the second specification adds two drinking indicators – an indicator of whether the respondent is a drinker and an indicator of whether the respondent is a frequent binger. Each coefficient reported in the table includes the z-statistic, the ratio of the estimated coefficient and standard error, which in the limit has a normal distribution. The logit regressions demonstrate that smoking is strongly (negatively) predictive of high school graduation and especially college matriculation for both males and females even after adjusting for family structure, socioeconomic status, academic ability, and other characteristics of the respondent. On the other hand, drinkers have about the same chances as nondrinkers
161
Smoke Signals: Adolescent Smoking and School Continuation
Table 1.
College Matriculation Rates for Smokers and Nonsmokers.
A. Males Percentage of Male Youth who Enrolled in a Four-Year College in the Year Immediately Following High School Graduation
By ASVAB math Scored below median (1.03) Scored at or above median (1.03) Mother’s education level Mom did not complete high school Mom completed high school Mom completed high school and some college Mom completed college By father’s education level Dad did not complete high school Dad completed high school Dad completed high school and some college Dad completed college By ethnicity Black Hispanic Other
Smoker during Junior Year of High School
Nonsmoker during Junior Year of High School
3.0% (199) 26.1% (138)
11.3% (592) 44.2% (651)
1.4% (71)
7.8% (256)
6.1% (131) 12.5% (104)
21.2% (524) 31.0% (329)
32.3% (62)
54.2% (286)
3.6% (56) 3.1% (97) 14.5% (69)
13.3% (181) 18.4% (337) 31.3% (243)
36.1% (61)
55.0% (300)
6.7% (60) 2.8% (72) 13.8% (276)
20.2% (392) 13.0% (322) 35.1% (823)
B. Females Percentage of Female Youth who Enrolled in a Four-Year College in the Year Immediately Following High School Graduation
By ASVAB math Scored below median (2.68) Scored at or above median (2.68) Mother’s education level Mom did not complete high school Mom completed high school Mom completed high school and some college Mom completed college By father’s education level Dad did not complete high school
Smoker during Junior Year of High School
Nonsmoker during Junior Year of High School
8.6% (209) 36.8% (163)
17.7% (600) 53.0% (647)
4.3% (70)
16.6% (271)
13.4% (134) 23.6% (127)
27.8% (507) 35.0% (329)
41.4% (58)
63.9% (291)
10.5% (38)
18.4% (196)
162
PHILIP J. COOK AND REBECCA HUTCHINSON
Table 1. (Continued ) B. Females Percentage of Female Youth who Enrolled in a Four-Year College in the Year Immediately Following High School Graduation
Dad completed high school Dad completed high school and some college Dad completed college By ethnicity Black Hispanic Other
Smoker during Junior Year of High School
Nonsmoker during Junior Year of High School
16.2% (105) 25.0% (76)
35.0% (317) 43.5% (200)
34.4% (64)
59.0% (285)
14.6% (55) 13.9% (65) 22.0% (318)
30.5% (440) 18.3% (333) 42.4% (759)
Note: Number in parentheses indicates total number of smokers or nonsmokers. Smoker is defined to be those respondents who had at least one cigarette in the 30 days preceding the respondent’s interview. Source: NLSY97; limited to those respondents who were juniors in high school in 1997, 1998, or 1999.
when smoking is held constant. The exception is the small group of females who are frequent bingers, who have a reduced likelihood of school continuation. The predictive importance of smoking is suggested by the fact that the estimated effect on the log odds of college matriculation of smoking is larger than the effect of having a father who is a dropout rather than a college graduate (other things equal). While smoking status is a strong predictor of school continuation, the most powerful predictor is the math ASVAB score. The coefficient estimates and z-statistic are reported in Table 3. As an example, consider the estimated effect of moving up by one standard deviation (SD) on ASVAB score. For males, this would increase the odds of graduating from high school by a factor of 1.7, and the odds of matriculating by a factor of 3.4. For females, moving up by one SD on ASVAB would increase the odds of graduating from high school by a factor of 2.0, and increase the odds of matriculating by 2.8.2 Since the logit specification is multiplicative, the absolute effect of smoking on the odds of school continuation increases with ability. This fact is illustrated by Fig. 3, which depicts the effect of being a smoker in junior year on the probability of college matriculation the following year over a wide range of ASVAB scores (four SDs). At two SDs below the mean the
163
Smoke Signals: Adolescent Smoking and School Continuation
Table 2.
High School Graduation Rates for Smokers and Nonsmokers. A. Males Percentage of Male Youth who Graduated High School by Age 19
By ASVAB math Scored below median (1.03) Scored at or above median (1.03) Mother’s education level Mom did not complete high school Mom completed high school Mom completed high school and some college Mom completed college By father’s education level Dad did not complete high school Dad completed high school Dad completed high school and some college Dad completed college By ethnicity Black Hispanic Other
Smoker during Junior Year of High School
Nonsmoker during Junior Year of High School
48.7% (199) 74.6% (138)
62.3% (592) 85.3% (651)
42.3% (71)
58.2% (256)
55.7% (131) 59.6% (104)
70.8% (524) 76.0% (329)
77.4% (62)
85.3% (286)
39.3% (56) 64.9% (97) 53.6% (69)
66.9% (181) 72.4% (337) 77.8% (243)
77.0% (61)
87.0% (300)
45.0% (60) 48.6% (72) 62.0% (276)
61.0% (392) 64.3% (322) 78.7% (823)
B. Females Percentage of Female Youth who Graduated High School by Age 19
By ASVAB math Scored below median (1.03) Scored at or above median (1.03) Mother’s education level Mom did not complete high school Mom completed high school Mom completed high school and some college Mom completed college By father’s education level Dad did not complete high school Dad completed high school
Smoker during Junior Year of High School
Nonsmoker during Junior Year of High School
65.1% (209) 84.7% (163)
72.3% (600) 89.5% (647)
52.9% (70)
67.5% (271)
69.4% (134) 77.2% (127)
76.1% (507) 82.1% (329)
84.5% (58)
90.4% (291)
63.2% (38) 66.7% (105)
71.4% (196) 78.9% (317)
164
PHILIP J. COOK AND REBECCA HUTCHINSON
Table 2. (Continued ) B. Females Percentage of Female Youth who Graduated High School by Age 19
Dad completed high school and some college Dad completed college By ethnicity Black Hispanic Other
Smoker during Junior Year of High School
Nonsmoker during Junior Year of High School
80.3% (76)
85.0% (200)
87.5% (64)
87.7% (285)
80.0% (55) 72.3% (65) 69.2% (318)
73.9% (440) 70.3% (333) 83.3% (759)
Note: Number in parentheses indicates total number of smokers or nonsmokers. Smoker is defined to be those respondents who had at least one cigarette in the 30 days preceding the respondent’s interview. Source: NLSY97; limited to those respondents who were juniors in high school in 1997, 1998, or 1999. ***Excludes missing observations.
likelihood of college matriculation is very small for both smokers and nonsmokers. Even though smoking is a negative signal, it is of little consequence given low ability. Above the mean the smoking gap becomes wide.
ALTERNATIVE PERSPECTIVES ON ADOLESCENT SMOKING Given that smoking is negatively linked to school continuation among the 11th graders, we expect that 11th graders will take this fact into account when deciding whether to smoke. Which of the following is important in a 16- or 17-year-old deciding whether to smoke? Tastes: the intrinsic pleasure of inhaling smoke and blowing it out again Price and availability of cigarettes and related commodities Concern about the possibility of becoming habituated to tobacco and someday suffering adverse health consequences Assessment of what peers and adults will think. The traditional economics literature has focused on the first three concerns. Tastes and price are considerations for any commodity. The fact that smoking has health consequences suggests that the decision to smoke is part
Smoke Signals: Adolescent Smoking and School Continuation
165
of the portfolio of investments and disinvestments in ‘‘health capital’’ (Grossman, 1972), and for this reason preferences concerning time and risk also become important. Because smoking is addictive, there is an additional set of delayed consequences associated with smoking including tolerance and habituation; these consequences may also influence current decisions (Becker & Murphy, 1988; Chaloupka, 1991; Becker, Grossman, & Murphy, 1991). Some evidence from this analysis points to a conclusion that youthful smoking decisions reflect a higher rate of time discount – greater ‘‘myopia’’ – than for adults (Chaloupka, 1991). A series of studies by Joni Hersch and her associates have produced evidence that adult smokers as a group tend to be risk takers, careless of their health and life in domains other than smoking (Hersch & Viscusi, 1990, 1998; Hersch & Pickton, 1995). ‘‘In terms of the level of risk, smokers are less likely to perform preventive health activities such as seatbelt use, flossing, and checking their blood pressure. They choose riskier jobs, are more likely to be injured on their jobs controlling for objective measures of risk, are more likely to have an accident at home, and are more likely to have an accident overall’’ (Hersch & Viscusi, 1998). Following this line of logic, an explanation for the negative link between smoking and school continuation among adolescents is that both are influenced by the time preferences of the individuals – other things equal, youths who are more present-oriented will be more likely than their peers to smoke (because they discount long-term negative consequences) and more likely to end their schooling early. But if youths care about the opinions of their peers, then the logic is not so clear. For 11th graders, smoking, as demonstrated in the previous section, is a signal of being off track and unsuccessful in the school domain. This message may cause smokers to lose standing with peers, most of whom aspire to graduate from high school and obtain a college education (Cook & Ludwig, 1997). This loss of standing will be an immediate cost and hence of particular salience to present-oriented youths. On the other hand, the cost of being identified as ‘‘off track’’ is less for those who are in fact off track, than for those who are not. For one thing, classmates will probably have reason to suspect the truth already from other behaviors, and in any event it will soon be revealed when the individual does terminate schooling. Part of the ‘‘signaling’’ story, then, is that those who are disaffected from school and inclined to terminate will find smoking more attractive than those who intend to continue, because the ‘‘off track’’ identity signal is less costly in terms of peer standing than for those who are on track. But there is more to the story, a positive benefit as well as a reduced cost. In popular
166
PHILIP J. COOK AND REBECCA HUTCHINSON A. Males
ASVAB Scores Score below median Score above median Mother's Education Did not complete HS Completed high school Completed some college Completed college Father's Education Did not complete HS Completed high school Completed some college Completed college Race Black Hispanic Other 0
20
40 Male Smoker
60
80
100
Male Nonsmoker
Percent Graduating High School
Fig. 1.
Probability of High School Graduation for Different Groups of 11th Graders.
culture, not to mention tobacco advertising, smoking has long been associated with being ‘‘cool,’’ independent, sophisticated, and rebellious – all values that are prized by adolescents, and which may get them off track in their school careers (Albaum, Baker, Hozier Jr., & Rogers, 2002; AloiseYoung & Hennigan, 1996; Luke et al., 2001; Kobus, 2003). Many teens valued the smoking image enough to wear promotional items from the cigarette companies when such items were permitted (Wakefield, Flay, Nichter, & Giovino, 2003). But while these associations are generally attractive to adolescents, they are disinclined to project an image or claim an identity that is too far out of line with reality (McKennell & Bynner, 1969; Chassin, Presson, Sherman, Coity, & Olshavsky, 1981; Grube, Weir, Getzlaf, &
167
Smoke Signals: Adolescent Smoking and School Continuation B. Females ASVAB Scores Score below median Score above median
Mother's Education Did not complete HS Completed high school Completed some college Completed college
Father's Education Did not complete HS Completed high school Completed some college Completed college
Race Black Hispanic Other 0
20
40 Female Smoker
60
80
100
Female Nonsmoker
Percent Graduating High School
Fig. 1.
(Continued)
Rokeach, 1984; Chassin, Presson, Sherman, McLaughlin, & Gioia, 1985; Burton et al., 1989). The psychologists have established that identification with the images associated with smoking is a strong predictor of becoming a smoker (Aloise-Young & Hennigan, 1996). In sum, those youths who smoke and thereby assume the image associated with being a smoker incur the reputational costs of being considered off track and the benefits of being associated with an activity that connotes cool, sophistication, and rebelliousness. The cost is less for those who are genuinely off track, and the image of cool detachment or rebellion is likely to be more appealing and successful for those who genuinely identify with it.3
168
PHILIP J. COOK AND REBECCA HUTCHINSON A. Males ASVAB Scores Score below median Score above median Mother's Education Did not complete HS Completed high school Completed some college Completed college Father's Education Did not complete HS Completed high school Completed some college Completed college Race Black Hispanic Other 0
20
40 Male Smoker
60
80
100
Male Nonsmoker
Percent Enrolled in College
Fig. 2.
Probability of College Matriculation for Different Groups of 11th Graders.
Finally, there is good reason to believe that the reputational costs of smoking are less for students of low ability than for those who are above average, simply because, as we have seen, the ‘‘smoking gap’’ in the likelihood of school success is so much larger for high-ability students.4 Bringing together these conceptual strands, we propose that the smoking decision by adolescents is made in a social context where being a smoker conveys a signal about their personality (‘‘cool’’) and their likely success in school (questionable). The smoking decision is made in circumstances where their peers may observe other signals of schooling success; in particular their academic ability is somewhat visible, and for this reason the negative connotations of the signal – its cost – are inversely related to academic ability. The cost of the signal also depends on whether they are in fact off
169
Smoke Signals: Adolescent Smoking and School Continuation B. Females ASVAB Scores Score below median Score above median Mother's Education Did not complete HS Completed high school Completed some college Completed college Father's Education Did not complete HS Completed high school Completed some college Completed college Race Black Hispanic Other 0
20
40
Female Smoker
60
80
100
Female Nonsmoker
Percent Enrolled in College
Fig. 2.
(Continued)
track; those who are off track and have embraced an oppositional identity will find it less costly (or perhaps even rewarding) to signal that identity. To translate this framework into symbols, we have: V ¼ student’s intrinsic taste for smoking C ¼ social cost of smoking to the student (positive or negative) A ¼ aptitude P ¼ extent to which the student is ‘‘on track’’ (committed to continuing school) F ¼ financial cost of smoking H ¼ current valuation of smoking cost to present and future health
170
PHILIP J. COOK AND REBECCA HUTCHINSON
Table 3.
Logit Results on High School Graduation and College Matriculation, Full Sample.
High School Graduation Logit Regression Results, Coefficient Estimates and z-statistic, All Races Smoker Drinker Frequent binger ASVAB Selected Eexplanatory Variables Males N ¼ 1,945 Specification 1 0.667 Specification 2 0.649 Females N ¼ 1,970 Specification 1 0.322 Specification 2 0.256
[5.2] [4.7]
0.081 [0.6]
0.297 [1.4]
0.0570 [7.6] 0.0564 [7.5]
[2.3] [1.6]
0.025 [0.2]
0.553 [2.0]
0.0754 [9.0] 0.0754 [9.0]
College Matriculation Logit Regression Results, Coefficient Estimates and z-statistic, All Races Selected Explanatory Variables Males N ¼ 1,945 Specification 1 1.149 [5.9] 0.1321 [12.2] Specification 2 1.168 [5.5] 0.040 [0.3] 0.004 [0] 0.1323 [12.1] Females N ¼ 1,963 Specification 1 0.670 [4.5] 0.1144 [12.2] Specification 2 0.563 [3.4] 0.033 [0.2] 0.706 [2.0] 0.1143 [12.1] Note: All regressions include, in addition to the regressors shown, the following: indicators of respondent’s race, mother’s and father’s education, household size, indicator of whether the family was intact, Ln of household income, urban residence, regional indicators, cohort indicators, and indicators for month of interview.
r ¼ rate of time discount s ¼ the strength of the statistical link between smoking and likelihood of school continuation V ¼ V(r), with V0 >0 (higher rate of time discount enhances valuation of smoking) C ¼ C(A, P: s), increasing in both A and P conditional on a negative association s P ¼ P(r), with P0 o0 (commitment to school declines as discount rate increases) H ¼ H(r), with H0 o0 (concern about health consequences declines as discount rate increases) The student will choose to smoke if V 4C½A; PðrÞ : s þ F þ HðrÞ
171
Smoke Signals: Adolescent Smoking and School Continuation 0.9 0.8 0.7
Probability
0.6 0.5 0.4 0.3 0.2 0.1 0 -2 -1.8 -1.6 -1.4 -1.2
-1 -0.8 -0.6 -0.4 -0.2
0
0.2 0.4 0.6 0.8
1
1.2 1.4 1.6 1.8
2
ASVAB score (standard units) Nonsmokers
Fig. 3.
Smokers
Probability of College Matriculation for Male 11th Graders.
Assuming that taste for smoking V is distributed in the student population independent of the other variables, then the likelihood a particular student will choose to smoke depends on aptitude (negatively) and commitment to school (negatively) to an extent that is influenced by the strength of the smoking signal; the likelihood of smoking is positively related to rate of time discount, both because of the salience of future health effects, and the commitment to continuing in school.
EMPIRICAL DETERMINANTS OF SMOKING AND DRINKING We estimated standard equations for smoking participation among the 11th graders in the NLSY sample, and have reported partial results in Table 4 (and complete results in Table A4). The unusual feature of the specifications is that they include indicators of subsequent events, namely, whether the respondent graduated from high school the following year and whether he matriculated in a four-year college. These indicators are intended as proxies for the extent to which the respondent is on track or off in 11th grade. One alternative is to utilize contemporaneous reports from the respondent about
172
PHILIP J. COOK AND REBECCA HUTCHINSON
Table 4. Smoking Logit Regression Results, Coefficient Estimates and z-statistic, All Races. Selected Explanatory Variables
Males N ¼ 1,894 Specification 1 Specification 2 Females N ¼ 1,934 Specification 1 Specification 2
High school graduate
College matriculant
ASVAB
% of peers who smoke
0.493 [3.7] 0.480 [3.6]
0.985 [5.1] 0.953 [4.9]
0.0216 [2.6] 0.0175 [2.1]
1.196 [4.7]
0.149 [1.0] 0.114 [0.8]
0.693 [4.5] 0.626 [4.0]
0.0237 [2.9] 0.0154 [1.8]
1.781 [7.1]
Note: See Table 3.
how far she expects to go in school (Gruber & Zinman, 2001), but such reports are not available in the NLSY. The main results come as no surprise given the previous findings. Those who are on track to graduating from high school are less likely to smoke, and those who are on track to matriculating in a four-year college are much less likely to smoke, in comparison to their classmates. This is true despite the fact that the specifications control for a number of variables, which may influence smoking, including socioeconomic status, race, family characteristics, and cigarette prices. We also find that ability, as measured by ASVAB, has a strong, independent negative effect on the likelihood of smoking, as suggested by the signaling story. The respondent’s report on the prevalence of smoking among her peers has a strong positive association with whether or not the respondent smokes.5 This may reflect peer influence, or it may be an artifact of the wellknown pattern that smokers choose to associate with other smokers and may get a biased impression of their overall prevalence as a result. In any event, it is interesting to note that controlling for peer smoking has little effect on the coefficients for high school graduation and college matriculation. For drinking, the influence of being on track is less consistently evident, as shown in Table 5. High school graduation is statistically irrelevant, while college matriculation has a modest negative effect on the likelihood of drinking. ASVAB has a weak effect. Interestingly, the percentage of peers who smoke is quite closely linked to the drinking decision. These results on smoking and drinking are compatible with both the ‘‘time discount’’ story and the ‘‘social cost’’ story for the smoking
Smoke Signals: Adolescent Smoking and School Continuation
Table 5.
173
Drinking Logit Regression Results, Coefficient Estimates and z-statistic, All Races. Selected Explanatory Variables
Males N ¼ 1,894 Specification 1 Specification 2 Females N ¼ 1,934 Specification 1 Specification 2
High school graduate
College matriculant
ASVAB
% of peers who smoke
0.086 [0.7] 0.078 [0.7]
0.345 [2.4] 0.327 [2.3]
0.0098 [1.4] 0.0077 [1.1]
0.555 [2.5]
0.058 [0.4] 0.036 [0.3]
0.338 [2.6] 0.300 [2.3]
0.0069 [0.9] 0.0120 [1.6]
0.966 [4.4]
Note: All regressions include, in addition to the regressors shown, the following: state cigarette tax, race indicators, indicators of mother’s and father’s education, household size, indicator of whether the family was intact, Ln of household income, urban residence, regional indicators, cohort indicators, and indicators for month of interview.
decision – and indeed, we believe they both have some validity. The fact that ability has a strong independent effect on smoking (even after controlling for whether the student is on track) is better captured by the ‘‘signaling’’ story than the ‘‘rate of time discount’’ story. The fact that being on track to high school graduation has no effect on whether the 11th grade respondent drinks may also be relevant in distinguishing between the two stories. Drinking is an activity that has much in common with smoking for an 11th grader – it is illegal, frowned on by authorities, and has the potential to do harm over time through addiction and direct risks to health and safety. (Recent research suggests that drinking can do permanent damage to the developing brain of an adolescent.) It would seem, then, that if time preference influenced the likelihood of smoking, it should also influence the likelihood of drinking. The lack of a finding in this respect adds credence to the second ‘‘signaling’’ perspective on smoking.
CONCLUDING THOUGHTS Our analysis has documented the strength of smoking as a signal of being off track for juniors in high school. This signal conveys much more information for youths who are likely to graduate and go on to college, than for those who are likely to exit early from schooling. If we make the plausible
174
PHILIP J. COOK AND REBECCA HUTCHINSON
assumption that adolescents care about what their peers believe about them, and that most youths do not want to be viewed by their peers as ‘‘losers,’’ then smoking has a social cost. We have demonstrated that this cost is lower for lower-ability students, and is likely to be lower for those who are genuinely disaffected and off track. These observations provide an explanation for observed patterns of smoking among 11th graders, including the strong link to ability and school continuation. This explanation should not be viewed as a rival to the traditional ‘‘time horizon’’ explanation in economics. ‘‘Peer reputation’’ and ‘‘time horizon’’ are both plausible explanations. Our point is that the ‘‘time horizon’’ explanation is not complete, and in particular does not explain the relationship between smoking and ability. Our analysis does not speak to the question of whether smoking has a causal effect on schooling. For example, we can’t say whether an increase in the tobacco tax would increase college matriculation rates. However, if peer effects are important to the decision of whether to smoke, and the social processes described here result in a tight link between smoking and oppositional culture, then it is not implausible that a youth with a strong taste for tobacco might hang out with ‘‘bad company’’ just for companionship in his smoking. This association in turn may push him off track with respect to schooling. Finally, we note that our results on smoking and schooling may not apply to other times and places. To the extent that smoking patterns are influenced by peer-group concerns, they are self-perpetuating, with the possibility of several stable equilibria. It would be interesting to replicate this study using U.S. data from an earlier era, say the 1950s, when smoking among adults was much more prevalent and less closely linked to education than now, or to analyze smoking and schooling patterns for other wealthy nations where smoking may convey a more ambiguous social signal than in the United States today.
NOTES 1. These results are based on counts of respondents, and do not employ the sample weights. 2. We experimented a bit with the sample and specification. The estimated effect of ASVAB is almost identical in baseline equations (not shown) that have no drinking or smoking variables. Table A3 presents the results of the same regressions run for the sub-sample of respondents who are white or ‘‘other’’ races – neither black nor Hispanic. Isolating this group is of interest because there does appear to be a strong ‘‘race’’ effect on smoking and drinking. As it turns out, however, the prediction equations are quite similar with respect to the key coefficients. We also experimented
Smoke Signals: Adolescent Smoking and School Continuation
175
with adding an interaction term (ASVAB* Smoking) to the prediction equations, but none of the coefficient estimates on this interaction term had a statistically discernible effect. (Most of the z-statistics were less than one.) 3. This story may be developed somewhat differently based on Akerlof and Kranton’s theory of identity and economic behavior (Akerlof & Kranton, 2000, 2002). They note that the school social scene is highly structured into groups or social categories, such as the leading crowd, nerds, and burnouts. Students have certain attributes that determine the utility payoffs from membership in each group, and then select one accordingly. Complying with the behavioral norms of the group has an immediate payoff from identification with that group even when the behavior conflicts with other interests of the individual. If smoking is a valued behavior among the burnouts, then those who select the burnout category for other reasons (a lack of interest in or ability for schoolwork) may begin smoking. 4. An interesting possibility is that the members of the student body who are most clearly destined for success will choose to smoke as a ‘‘counter signal’’ that separates them from the middle range of on-track students. In an article with the perfect title ‘‘Too cool for school?,’’ Feltovich, Harbaugh, and To note that ‘‘high types sometimes avoid the signals that should separate them from lower types, while intermediate types often appear the most anxious to send the ‘right’ signals. The nouveau riche flaunt their wealth, but the old rich scorn such gauche displays y Mediocre students answer a teacher’s easy questions, but the best students are embarrassed to prove their knowledge of trivial points’’ (Feltovich, Harbaugh, & To, 2002, p. 631). This effort by the elite to separate from the middle is possible, they note, in a circumstance where there are other signals of success. 5. For the peer-smoking question, respondents are shown a card that has five choices, and told to choose one: 1. Almost none (o10%); 2. About 25%; 3. About half (50%); 4. About 75%; 5. Almost all (>90%).
REFERENCES Albaum, G., Baker, K. G., Hozier, G. C., Jr., & Rogers, R. D. (2002). Smoking behavior, information sources, and consumption values of teenagers: Implications for public policy and other intervention failures. Journal of Consumer Affairs, 36(1), 50–76. Akerlof, G. A., & Kranton, R. E. (2000). Economics and identity. Quarterly Journal of Economics, 105(3), 715–753. Akerlof, G. A., & Kranton, R. E. (2002). Identity and schooling: Some lessons for the economics of education. Journal of Economic Literature, 40(4), 1167–1201. Aloise-Young, P. A., & Hennigan, K. M. (1996). Self-image, the smoker stereotype and cigarette smoking: Developmental patterns from fifth to eighth grade. Journal of Adolescence, 163–177. Banfield, E. (1974). The unheavenly city revisited. Bsoton: Little, Brown and Company. Banfield, E. (1977). Present-orientedness and crime. In: R. E. Barnett & J. Hagel (Eds), Assessing the criminal: Restitution, retribution, and the legal process. Cambridge: Ballinger. Becker, G. S., Grossman, M., & Murphy, K. M. (1991). Rational addiction and the effect of price on consumption. American Economic Review, 81(2), 237–241. Becker, G. S., & Murphy, K. M. (1988). A theory of rational addiction. Journal of Political Economy, 96(4), 675–700.
176
PHILIP J. COOK AND REBECCA HUTCHINSON
Bo¨hm-Bawerk., Eugen von. ([1884–1921] 1959). Capital and interest, Kapital und Kapitalzins, South Holland, IL: Libertarian Press. Burton, D., et al. (1989). ImAge attribution and smoking intentions among seventh grade students. Journal of Applied Social Psychology, 19, 656–664. Center for Human Resource Research. (2002). NLSY97 User’s Guide. Columbus, OH: Ohio State University. Chaloupka, F. (1991). Rational addictive behavior and cigarette smoking. Journal of Political Economy, 99(4), 722–742. Chassin, L., Presson, C. C., Sherman, S. J., Coity, E., & Olshavsky, R. W. (1981). Self-images and smoking in adolescence. Personality and Social Psychology Bulletin, 7, 670–676. Chassin, L., Presson, C. C., Sherman, S. J., McLaughlin, L., & Gioia, D. (1985). Psychosocial correlates of adolescent smokless tobacco use. Addictive Behaviors, 10, 431–435. Cook, P. J., & Ludwig, J. (1997). Weighing the burden of ‘acting white’; Are there race differences in attitudes towards education?. Journal of Policy Analysis and Management, 16(2), 256–278. Feltovich, N., Harbaugh, R., & To, T. (2002). To cool for school? Signalling and countersignalling. RAND Journal of Economics, 33(4), 630–649. Grossman, M. (1972). On the concept of health capital and the demand for health. Journal of Political Economy, 80, 223–255. Grube, J. W., Weir, I. L., Getzlaf, S., & Rokeach, M. (1984). Own value system, value images and cigarette smoking. Personality and Social Psychology Bulletin, 10, 306–313. Gruber, J., & Zinman, J. (2001). Youth smoking in the United States: Evidence and implications. In: J. Gruber (Ed.), Risky behavior among youths: An economic analysis (pp. 69–120). Chicago: University of Chicago Press. Hersch, J., & Pickton, T. S. (1995). Risk-taking activities and heterogeneity of job-risk tradeoffs. Journal of Risk and Uncertainty, 11(3), 205–217. Hersch, J., & Viscusi, W. K. (1990). Cigarette smoking, seatbelt use, and differences in wagerisk tradeoffs. Journal of Human Resources, 25(2), 202–227. Hersch, J., & Viscusi, W. K. (1998). Smoking and other risky behaviors. Journal of Drug Issues, 28(3), 645–661. Hoppe, H. (2001). Democracy-the God that failed. New Brunswick, NJ: Transaction Publishers. Kobus, K. (2003). Peers and adolescent smoking. Addiction, 98(Suppl 1), 37–55. Luke, D., et al. (2001). Teen’s images of smoking and smokers. Public Health Reports, 116, 194–202. McKennell, A. C., & Bynner, J. M. (1969). Self images and smoking behavior among school boys. British Journal of Educational Psychology, 39, 27–39. Menger, C. ([1871] 1976). Principles of economics, Grundsa¨tze der Volkswirtschaftslehre. New York: New York University Press. Mises, L. ([1949] 1998). Human action: A treatise on economics (5th ed). Auburn: Ludwig von Mises Institute. Roberts, R. D., et al. (2000). The armed services vocational aptitude battery. Learning and individual differences, 12(1), 81–103. Rothbard, M. N. ([2004] 1963). Man, economy, and state (2nd ed). Auburn, MI: Ludwig von Mises Institute. Smith, T. A. (1988). Time and public policy. Knoxville: University of Tennessee Press. Strigl, R. (2001). Capital and production. Auburn, MI: Ludwig von Mises Institute. Wakefield, M., Flay, B. R., Nichter, M., & Giovino, G. (2003). Role of the media in influencing trajectories of youth smoking. Addiction, 98(Suppl 1), 79–103. Wilson, J. Q., & Herrnstein, R. J. (1985). Crime and human nature. New York: Simon & Schuster.
Table A1. Variables
Descriptive Statistics, Variable Definitions, and Sample Construction. Mean (Standard Deviation) Males
Definitions
Females
Educational attainment High school graduate
0.7123 (0.4512)
0.7745 (0.4194)
Enrolled in college
0.2579 (0.4360)
0.3301 (0.4689)
Smoking/drinking habits Junior year smoker
0.2243 (0.4158)
0.2586 (0.4567)
Junior year drinker
0.3702 (0.4812)
0.3559 (0.4805)
Missing junior year drinker
0.0254 (0.1567)
0.0182 (0.1341)
Junior year frequent binge drinker
0.1107 (0.3122)
0.0662 (0.2500)
Education d.v.: 1 if youth graduated high school by age 19; 0 otherwise Education d.v.: enrolled in a fouryear college in the year immediately following high school graduation
177
Smoking d.v.: 1 if youth reported smoking at least one or more cigarettes per days in the past 30 days during junior of high school Drinking d.v.: 1 if youth reports any drinking in the past 30 days during junior year of high school Missing Data d.v.: 1 if drinking habits were not reported Drinking d.v.: 1 if youth reports drinking on 5 or more drinks on 4 or more occasions during the past 30 days during junior year of high school
Smoke Signals: Adolescent Smoking and School Continuation
APPENDIX
178
Table A1 (Continued ). Variables
Missing junior year frequent binge drinker
Race & ethnicity Black Hispanic Parent’s education levels Mother did not finish high school Mother finished high school
Mother completed some college Mother finished college
Definitions
Males
Females
0.2098 (0.4059)
0.2528 (0.4361)
40.90 (23.77) 0.0337 (0.1798)
38.40 (24.34) 0.0185 (0.1549)
0.4834 (0.0852) 0.0263 (0.1594)
0.4883 (0.0902) 0.0185 (0.1353)
0.1339 (0.3394) 0.1311 (0.3364)
0.1529 (0.3612) 0.1157 (0.3210)
Race d.v.: 1 if black Race d.v.: 1 if Hispanic
0.1380 (0.3441)
0.1392 (0.3470)
0.3770 (0.4835)
0.3564 (0.4802)
0.2638 (0.4397)
0.2754 (0.4464)
0.2212 (0.4141)
0.2290 (0.4213)
Parent’s education d.v.: 1 if mother did not complete high school Parent’s Education d.v.: 1 if Mother completed high school but did not attend any college Parent’s Education d.v.: 1 if Mother completed some college Parent’s Education d.v.: 1 if Mother completed college
Missing Data d.v.: 1 if drinking habits were not reported, 0 otherwise
Missing Data d.v.: 1 if tax index was not reported Missing Data d.v.: 1 if tax index was not reported
PHILIP J. COOK AND REBECCA HUTCHINSON
State level indicators State cigarette tax index Missing state cigarette tax index State beer tax index Missing state beer tax index
Mean (Standard Deviation)
Father completed some college Father finished college
0.0883 (0.2828)
0.0840 (0.2784)
0.1416 (0.3490)
0.1411 (0.3479)
0.3200 (0.4669)
0.3318 (0.4706)
0.2379 (0.4262)
0.2184 (0.4129)
0.3004 (0.4589)
0.3088 (0.4617)
Missing father’s education level Family & household Household size
0.2598 (0.4371)
0.2918 (0.4562)
4.3217 (1.3660)
4.2944 (1.4820)
Missing household size
0.0246 (0.1543)
0.0165 (0.1278)
Family intact in 1997
0.7695 (0.4197)
0.7278 (0.4467)
Missing family intact in 1997 Household income (Ln) Income missing
–
0.0018 (0.0424)
10.6728 (0.9102)
10.6222 (1.0256)
0.7475 (0.4330)
0.7014 (0.4592)
Missing Data d.v.: 1 if Mother’s education level was not reported Parent’s Education d.v.: 1 if Father did not complete high school Parent’s Education d.v.: 1 if Father completed high school but did not attend any college Parent’s Education d.v.: 1 if Father completed some college Parent’s Education d.v.: 1 if Father completed college Missing Data d.v.: 1 if Father education level was not reported Number of persons living in Household with youth . Missing data d.v.: 1 if household size was not reported Family d.v. 1 if youth lived with both biological, step, or adopted parents in 1997 Missing data d.v.: 1 if family intact data was not reported Log of household income reported by youth. Missing data d.v.: 1 if income data was not reported
Smoke Signals: Adolescent Smoking and School Continuation
Missing mother’s education level Father did not finish high school Father finished high school
179
180
Table A1 (Continued ). Variables
Mean (Standard Deviation) Males
Definitions
Females
1.9771 (9.2245)
3.2417 (8.9102)
Missing math ASVAB
0.1730 (0.3770)
0.1635 (0.3711)
ASVAB Math Aptitude Score, divided by 100. Scores were scaled such the median was ‘‘0’’ Missing Data d.v.: 1 if Math ASVAB score was not reported
Peers Percentages of respondent’s peer who smoke Demographics Urbanity
0.3364 (0.2367)
0.4038 (0.2534)
Self-reported fraction of peers that respondents report as smokers.
0.7192 (0.4484)
0.7072 (0.4562)
Missing urban
0.0566 (0.2302)
0.0474 (0.2131)
Body weight Missing weight
162.4 (34.0) 0.0298 (0.1696)
133.2 (27.5) 0.0401 (0.1970)
South
0.2996 (0.4567)
0.3394 (0.4750)
Demographic d.v.: 1 if youth reports living in an urban region Missing Data d.v.: 1 if no urban/rural information is reported Youth’s body weight in pounds Missing Data d.v.: 1 if no weight was reported Demographic d.v.: 1 if youth lives in South Census Region
West Northeast North central Missing census region
0.2258 0.1911 0.2817 0.0246
0.2126 0.1795 0.2671 0.0164
(0.4169) (0.3920) (0.4485) (0.1543)
(0.4104) (0.3850) (0.4439) (0.1275)
PHILIP J. COOK AND REBECCA HUTCHINSON
Aptitude Math ASVAB score
0.3196 (0.0727)
0.3535 (0.4797)
Cohort 2
0.3621 (0.4790)
0.3324 (0.4727)
Cohort 3
0.3183 (0.4643)
0.3141 (0.4658)
Interaction terms Month of interview January February March April May June July August September October November December Number of observations
0.1158 (0.3190) 0.1039 0.0997 0.1051 0.0640 0.0391 0.0164 0.0099 0.0011 0.0188 0.2251 0.1765 1,945
0.1188 (0.3247) 0.1077 0.0953 0.1118 0.0749 0.0406 0.0191 0.0063 0.0025 0.0174 0.2094 0.1797 1,970
Sample d.v.: 1 if youth was a junior in high school in 1997 Sample d.v.: 1 if youth was a junior in high school in 1998 Sample d.v.: 1 if youth was a junior in high school in 1999
Month of interview d.v.: 1 if youth was interview of during January of junior year
Smoke Signals: Adolescent Smoking and School Continuation
Cohort 1
Note: All means & standard deviations are weighted and exclude missing observations. All figures are weighted by a sampling weight. Source: NLSY97; limited to those respondents who were juniors in high school in 1997, 1998, or 1999.
181
182
Table A2.
Educational Attainment: Association with Smoking in 11th Grade.
Graduated High School by Age 19 Males
Females
Enrolled in a four-year College the Year Following High School Graduation Males
0.3070c (0.1593) 0.9751a (0.2077) Smoker during the 11th 0.6822a (0.1436) grade Race and ethnicity 0.2444 (0.2120) Black 0.1482 (0.1745) 0.5525a (0.1984) 0.2264 (0.2413) Hispanic 0.2620 (0.1784) 0.3942c (0.2110) Parent’s education levels (Defaults: Mom is school dropout, Dad is school dropout) Mom finished high 0.2570 (0.1936) 0.2007 (0.2079) 0.7271b (0.3174) school Mom finished some 0.2742 (0.2133) 0.4251c (0.2339) 0.9196a (0.3246) college Mom finished college 0.5846b (0.2687) 0.8105a (0.3028) 1.2133a (0.3410)
Females 0.6404a (0.1623) 0.7390a (0.1819) 0.0328 (0.2041) 0.1078 (0.2369) 0.3825 (0.2482) 1.1179a (0.2666)
PHILIP J. COOK AND REBECCA HUTCHINSON
Logit Estimates, Coefficients, and Standard Errors
0.1799 (0.2239) 0.1328 (0.2481)
0.0639 (0.2486) 0.6629b (0.2988)
0.7008b (0.2863)
0.4091 (0.3085)
0.0476 (0.0449) 0.3306c (0.1828) 0.0045 (0.1408)
0.0350 (0.0465) 0.2484 (0.2114) 0.0803 (0.1111)
0.6311c (0.3238) 0.1068 (0.3223)
0.3649 (0.2623) 0.6282b (0.2785)
0.6354c (0.3274)
0.5923b (0.2821)
0.0185 (0.0581) 0.0182 (0.2464) 0.1608 (0.2078)
0.0905c (0.0487) 0.3030 (0.2216) 0.1751 (0.1277)
0.0575a (0.0078) 0.0785a (0.0088) 0.1370a (0.0113) 0.1177a (0.0097) 0.1354 0.1537 0.2919 0.2279 1580 1616 1578 1613
Note: Smoker is defined to be those respondents who had at least one cigarette in the 30 days preceding the respondent’s interview. Other variables included are indicators for urban residence, census region, body weight, cohorts (which year the respondent was a junior: 1997, 1998, or 1999), month of interview indicators as well as missing variable indicators for state level indicators, parent’s education levels, household income, ASVAB scores, urban residence, body weight, and census region. Source: NLSY97; limited to those respondents who were juniors in high school in 1997, 1998, or 1999. a Significantly different than zero at the 1% level. b Significantly different than zero at the 5% level. c Significantly different than zero at the 10% level.
Smoke Signals: Adolescent Smoking and School Continuation
Dad finished high school Dad finished some college Dad finished college Family & household Household size Family intact in 1997 Household income (Ln) Other Math ASVAB score R2 Number of observations
183
184
Table A3.
Logit Regression Results, Coefficient Estimates and z-statistic White, Asian and Other.
High School Graduation Smoker
Drinker
Frequent binger
ASVAB
0.705 [4.2] 0.607 [3.3]
0.117 [0.6]
0.741 [2.9]
0.0609 [5.7] 0.0593 [5.5]
.638 [3.6] 0.593 [3.0] Smoker
Males (N ¼ 1099) Smoking only Smoking and drinking Females (N ¼ 1072) Smoking Smoking and drinking
0.138 [0.7] 0.636 [1.9] Selected Covariates Drinker Frequent Binger
1.162 [5.2] 1.205 [4.9]
0.004 [0]
0.712 [4.0] 0.722 [3.6]
0.209 [1.2]
0.184 [0.6]
0.578 [1.5]
.0845 [6.7] 0.0845 [6.7] ASVAB 0.1505 [10.4] 0.1516 [10.3] 0.1294 [9.7] 0.1283 [9.6]
Note: All regressions include, in addition to the regressors shown, the following: indicators of mother’s and father’s education, household size, indicator of whether the family was intact, Ln of household income, urban residence, regional indicators, cohort indicators, and indicators for month of interview.
PHILIP J. COOK AND REBECCA HUTCHINSON
Males (N ¼ 1,099) Smoking only Smoking and drinking Females (N ¼ 1,074) Smoking Smoking and drinking College Matriculation
Selected Covariates
Smoking in 11th Grade: Association with Subsequent Schooling. Logit Estimates, Coefficients, and Standard Errors Males
Educational attainment High school graduate 0.6792a (0.1446) Enrolled in college – State level indicators State cigarette tax index 0.0061c (0.0035) Race and ethnicity Black 1.3769a (0.2096) Hispanic 0.4664b (0.1974) Parent’s education levels (Defaults: Mom is school
0.2815c (0.1590) – 0.0074b (0.0033)
Males – 0.9674a (0.2006) 0.0055 (0.0035)
Females – 0.6836a (0.1618) 0.0078b (0.0033)
1.5261a (0.2052) 1.1333a (0.2081) 0.7785a (0.2009) 0.4572b (0.1959) dropout, Dad is school dropout)
1.4933a (0.2056) 0.7881a (0.2007)
0.0257 (0.2181) 0.2200 (0.2347) 0.2121 (0.2752) 0.2221 (0.2457) 0.1038 (0.2669) 0.4960c (0.2892)
0.1511 (0.2166) 0.1165 (0.2280) 0.4866c (0.2704) 0.4971c (0.2615) 0.6030b (0.2802) 0.2153 (0.2927)
0.0390 0.2231 0.2673 0.3001 0.1101 0.4219
(0.2148) (0.2319) (0.2739) (0.2433) (0.2645) (0.2896)
0.1783 (0.2172) 0.1188 (0.2286) 0.3922 (0.2727) 0.5172b (0.2619) 0.6419b (0.2808) 0.2604 (0.2936)
0.0338 (0.0496) 0.2680 (0.2083) 0.0589 (0.1497)
0.1118b (0.0483) 0.6056a (0.2088) 0.0162 (0.1108)
0.0221 (0.0496) 0.3226 (0.2075) 0.0700 (0.1516)
0.1137b (0.0480) 0.6251a (0.2084) 0.0172 (0.1114)
185
Mom finished high school Mom finished some college Mom finished college Dad finished high school Dad finished some college Dad finished college Family and household Household size Family intact in 1997 Household income (Ln)
Females
Smoke Signals: Adolescent Smoking and School Continuation
Table A4.
186
Table A4 (Continued ). Logit Estimates, Coefficients, and Standard Errors Females
Males
Females
0.0363a (0.0083) 0.0710 1551
0.0335a (0.0083) 0.0883 1605
0.0301a (0.0085) 0.0734 1151
0.0257a (0.0085) 0.0972 1605
Note: Smoker is defined to be those respondents who had at least one cigarette in the 30 days preceding the respondent’s interview. Source: NLSY97; limited to those respondents who were juniors in high school in 1997, 1998, or 1999. a Significantly different than zero at the 1% level. b Significantly different than zero at the 5% level. c Significantly different than zero at the 10% level.
PHILIP J. COOK AND REBECCA HUTCHINSON
Other Math ASVAB score R2 Number of observations
Males
FASHION, GROWTH AND WELFARE: AN EVOLUTIONARY APPROACH Andreas Chai, Peter E. Earl and Jason Potts 1. INTRODUCTION The task of this paper is to explore the interplay between fashion, consumer lifestyles and economic growth in the context of a world of technological change in which the menu of possibilities that consumers face is constantly changing and tending to increase in length. Our working definition of ‘fashion’ is simple, namely the tendency or behavioural norm of actors to adopt certain types or styles of customs or commodities nearly simultaneously, only to adopt a different type or style of custom or commodity in future periods. The demand spikes associated with fashion may pertain to newly introduced products or to products that have been around for some time; they may also occur in hybrid cases where a seemingly defunct product or genre is given a brief rebirth by being reincarnated in terms of a new technology. Clearly, this is not a place for equilibrium analysis of the orthodox kind in which all that consumers are adjusting to are relative price changes and changes in their life-cycle stage that affect what they wish to consume. We need to understand how consumers cope with the problem of choice in the face of such changes and how their preferences change endogenously.1 A better understanding of the conditions under which novel goods emerge The Evolution of Consumption: Theories and Practices Advances in Austrian Economics, Volume 10, 187–207 Copyright r 2007 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 1529-2134/doi:10.1016/S1529-2134(07)10008-9
187
188
ANDREAS CHAI ET AL.
and how they are disseminated will have significant consequence for modelling macroeconomic growth. Central issues here – as with any evolutionary analysis – are the way and rate at which novel goods emerge and are adapted into the system (Dopfer, Foster, & Potts, 2004; Witt, 2003). If we add to the ingredients the fact that many of the goods that consumers can and do buy are very long-lived, then the challenges that the changing possibility set present become all the more significant. Obsolete knowledge makes errors likely, and once errors have been made in respect of durable items their consequences may be long-lived. Fashion may be frivolous in its drivers, but it may be durable in its consequence. Set against this backdrop, fashion appears problematic: we move from considering the scope for, and consequences of, individuals making costly errors to the possibility of people en masse making choices that they will later view with considerable embarrassment. The transitory popularity of products in the market place could, of course, simply be a reflection of a combination of their durability and functionality.2 Take, for example, DVD players and digital cameras. They were launched and people bought them en masse because they liked their enhanced functionality compared with VCRs and film-based cameras. Within a matter of years demand for them will tend to fall back to levels that reflect replacement needs (reliable designs may last 20 years or more) and growth associated with demographic change. In such cases, economists might be concerned about coordination problems associated with mass adoption insofar as production involved investment in assets that could not be used for the production of other things: a more gradual adoption profile would permit demand to be satisfied with less investment. But, otherwise, economists would not be particularly concerned: consumers will pay more than they would have had to do if adoption were more gradual, but their willingness to pay must reflect the value of the new product to them. Also, firms are often able to smooth out adoption spikes to some degree via price discrimination strategies. In the case of DVD players and digital cameras, the asset specificity issue may not be very significant, as much of the investment will be reused in next-generation products that offer even better performance. Moreover, the earlier generation products will tend to get reallocated within the household: a household’s first DVD player will probably get hooked up to a bedroom TV when a DVD recorder is bought, while a 3-megapixel camera might be given to a child when the parents invest in a digital SLR camera, and so on. Such examples, though not at odds with our definition of fashion, would normally tend to be placed under the heading of ‘product life cycle economics’ rather than fashion, for they have a clear rational choice basis in terms of functionality.
Fashion, Growth and Welfare: An Evolutionary Approach
189
The spikes in demand that seem to represent more cause for concern have a rather different form: these are cases where the product is not adopted for the long term and is abandoned long before it, or the specific assets required for its production are worn out. Here, the demand spikes are not driven by functionality but by style and social considerations. Whereas one can defend in terms of consumer time preference the tendency of firms to launch products that they fully expect to render obsolete once they have worked out how to make even better ones, it seems much harder to defend attempts of firms in fashion-dominated industries to induce obsolescence in terms of style. A superseded computer that is no longer being used because it cannot make use of broadband Internet access carries quite different connotations from a wardrobe stuffed with last year’s clothing still in as-new condition but not worn purely because it is now an embarrassment to wear. If fashion cycles involve waste, an ecological economist will naturally be concerned about them. But they ought to be of great interest to any economist who professes faith in market processes as means of enabling participants in the economy to get the most out of their resources. Within modern microeconomics, however, the theory of fashion has a curious but definite pariah status. It deals with phenomena that are too ephemeral and bourgeois for serious micro-economic theorists to touch and that are seemingly irrelevant for macro-economists to use as an explanatory variable in equations that aim to pin down the causes of growth. In short, fashion is a topic that to mainstream economists seems a bit too much like sociology or cultural studies, or worse, to warrant serious attention. Fashion is irrelevance multiplied by pretension, and therefore not a serious object of study. For the Austrians, fashion poses a particular challenge in relation to the social role of the entrepreneur. Within any fashion cycle, one can clearly see Kirznerian entrepreneurs making profits by being alert to potential or imminent spikes in demand or arbitraging between markets in which the timing of demand spikes differs. However, if, with the assistance of further entrepreneurial input, consumers keep turning their backs to the products on which the entrepreneurs have helped them spend their money, one must start questioning whether entrepreneurs in fashion-dominated sectors are performing a socially desirable function. To argue that entrepreneurs in these sectors are helping to serve a useful role rather than acting in a way that leads to more resources being wasted, either of two conditions needs to apply: (a) Consumers anticipate that their taste for the product will have a shorter life than the product itself and nonetheless still find it worth buying despite being aware that if they then dispose of it they will make a
190
ANDREAS CHAI ET AL.
capital loss on a scale far greater than one that merely reflected its physical depreciation; or (b) Despite consumers over-estimating the consumption benefits they derive from fashion goods, these losses are at least offset by social benefits that emerge as externalities of fashion cycles. In this paper, we focus on these two conditions and argue that fashion is integral to economic growth. Just as certain market dynamics are responsible for stimulating the (re)organization of production techniques of firms, so, too, market dynamics can be thought of as stimulating the (re)organization of consumer capabilities, which have important implications for understanding the nature and direction of economic evolution. We argue that fashion is a mechanism for periodically liquidating elements of consumer lifestyles in a world where there is a continual flow of novel consumer goods. Changes in fashion entail the mass updating of durable goods that work to control and accelerate the depreciation of existing goods, thereby lowering the mass adoption costs of new goods into consumers’ lifestyles. These adjustment costs are spread independently of whether consumers made good choices or bad choices in the previous rounds and they mitigate the effects of consumer mistakes. Fashion is to consumer theory in an evolving economic system what the liquidationist thesis of structural cleansing is to macroeconomics, under those same dynamic conditions (e.g. Caballero & Hammour, 1994).3 Our proposed evolutionary theory of fashion turns the standard view of fashion in microeconomics and consumer theory on its head. Instead of viewing fashion as a profligate bourgeois indulgence, we argue that it is an essential mechanism in the economic evolution of a market-capitalist system.
2. OLD-SCHOOL FASHION VERSUS A FOCUS ON NOVELTY The literature on fashion really begins with a remarkable paper by Caroline Foley (1893), but most modern articles in economics that link fashion and economic theory take Thorstein Veblen (1899) as defining the state of the art. Veblen, typical of high-powered intellectual outsiders, was very much down on fashion. Indeed, The Theory of the Leisure Class (1899) reads like an analysis of a virulent social pathology. The centrepiece of Veblenian microeconomics is the theory of conspicuous consumption, which seeks to explain changing consumption patterns from the agent’s basic desire for
Fashion, Growth and Welfare: An Evolutionary Approach
191
social status. Fashion is in this sense a kind of higher-order consumption, driven not by innate utility, but by its effects on other agents. According to Veblen, agents value and choose goods for status competition by emulating the tastes of other individuals situated at higher points in the social hierarchy. Where one is situated in the social hierarchy is decided by income but, according to Veblen’s analysis of late-nineteenth century American nouveaux riches, income alone does not equate to status. Key to transforming wealth into status is the social performance of the individual in terms of conspicuous consumption. Status derives from the judgments that other members of society make of an individual’s position in society. For this position to be established there must be a display of wealth, i.e. conspicuous consumption. As the lower classes imitate the higher classes, the higher classes must come up with more conspicuous and wasteful ways to display their wealth. As such, Veblen viewed fashion as symptomatic of the inherent instability of the market-capitalist system. Veblen’s disapproving view of fashion runs into trouble as soon as we try to reconcile his idea of conspicuous with Lancaster’s (1966) view that the demand for novelty can also be understood in terms of the consumer search for potential improvement in a commodity’s functional properties. For improved functionality of products to be saleable to status-hungry consumers, the fact that they are consuming the latest generation of products must be conspicuous, even if their improved functionality is ‘under the skin’. Otherwise, the status-seekers’ expenditure will tend to go to products that are more cost-effective at signalling that one can afford the latest generation. Hence, firms dominated by engineers who are desperate to compete by adding improved functionality need also to spend on re-skinning their products even if there is no functional need to do so. For example, adding safety features to cars is an ‘under-the-skin’ activity. Clearly, manufacturers can spend on advertising such additions and in the early days of anti-lock braking systems (ABS) and airbags being fitted to mass-market products they posted exterior signs such as rear badges that proclaimed ABS and door pillar signs that indicated (to those who knew the jargon) the presence of an airbag with the initials SRS. However, much more conspicuous cosmetic changes such as new lights, bumpers, wheels and trim garnishing may be far better effective ways of ensuring customers will buy vehicles with enhanced safety features – if indeed firms find it profitable to add the latter rather than just concentrating on the former. Note, too, that in this case, the kind of safety engineering that it will pay to incorporate may be affected by the extent to which it can be brought to the surface with a recognizable symbol. We are thus not at all surprised to recall that the first
192
ANDREAS CHAI ET AL.
Australian cars to be fitted with ABS and airbags, and their accompanying badges, in the mid-1990s continued to score poorly in crash-testing despite being sold with a new emphasis on safety, because their passenger compartments still crumpled badly. A decade after Lancaster’s attempt to get economists to understand the process of change in the technology of consumption, Tibor Scitovsky (1976) made some progress in understanding the economic consequences of consumer demand for novelty. Scitovsky relied on previous studies in neuropsychology that postulated that novel experiences stimulate changes in a person’s arousal levels that, if in the right direction, lead to the sensation of pleasure. Scitovsky argued that in reaching historically high levels of comfort in their lives, modern consumers paradoxically decrease their levels of pleasure, which derive from changes in comfort levels. The search for stimulation is the search for novelty. Fashion is the touchstone of social novelty. Thus, consumers demand novelty to attain utility and fashion becomes quasi-rational. This argument, together with that of Lancaster, provides a basis on which fashion trends can be understood as the coordinated introduction of novelty into society. It is this coordination-focused view that we think makes sense of the nature of fashion in an economic system in terms of the evolutionary dynamics of growth which are driven by the novel consumer good and the status-seeking behaviour of agents who consume goods socially. If we can understand consumers demanding novelty for the sake of arousal on the individual level, then it is simple to see how status competition or the use of fashion goods more broadly as tools of social communication works in the context of gaining the attention of other members of society through consuming new items whose very novelty makes them attention arousing. As such, they invite onlookers to make judgments about the quality of choice that the status-seeker has made and hence about the status to which the status-seeker is due. This is much more in keeping with the anthropological analysis of consumption proposed by Douglas and Isherwood (1978), which gives a much broader view of its social side than Veblen offers: people consume to communicate, and there is much more to communicate by being up with the fashions than merely one’s income or wealth. If the way that we adapt to new fashions signals something about our competences and how we see the world, it thereby helps social coordination. From Veblen’s standpoint, the concern of consumers with fashion is a social phenomenon. The fashion-conscious consumer is demanding a social good, namely status, which is the ‘dominant feature in the scheme of life’ (Veblen, 1899) and fashion would not exist if there were no one to impress.
Fashion, Growth and Welfare: An Evolutionary Approach
193
From Scitovsky’s standpoint, by contrast, fashion is an artefact of the individual’s demand for novelty; his argument does not ultimately depend on the existence of society, but rather on the fleeting nature of the individual’s attention and the satisfaction of desire for stimulation in the form of novelty and the new stimulus and relational structures this brings. Fashion, in this view, is all about satisfying one’s curiosity and seeking stimulating experiences. From this argument we understand why ultimately consumption items must be replaced, given the intrinsically fleeting nature of novelty. In both cases, what is demanded, produced and exchanged is attention or stimulus. Combining these two theories permits an understanding of how one can view consumption decisions as investment decisions without necessarily focusing upon the physical durability of the things which consumers buy. As with the purchase of consumer durables, expenditure on services and nondurable goods (for example, fitness club membership or meals in a particular restaurant, respectively) may be undertaken as an investment in building one’s social standing. Each of these kinds of spending involves risk because fashion goods are what Nelson (1970) would label as ‘experience goods’: novel aspects of utility-yielding properties cannot be assessed in advance via search, while the social response to an act of consumption can only be conjectured at the time of purchase. The possibility of vicarious learning from the experimentation of others with novel products means there is no need for a Veblenian or anthropological perspective on consumption for it to have a significant social side in a world in which consumption possibilities constantly change. Now, if fashion consumption is done for both individual and social reasons, then the obvious question that follows is about how these two distinct motivations interact to produce aggregate fashion consumption patterns. And this, we argue, requires a theory of how consumers strategically manage their consumption in a turbulent, changing and uncertain environment. Understanding a class of consumption decisions as investment decisions is one thing, but analysing the environment and the fluctuating determinants of these decisions is something altogether different.
3. CONSUMPTION COMPLEMENTARITIES AND MISTAKES Evolutionary economists argue that the growth of knowledge drives the growth of economic systems (Loasby, 1999). Choice is not so much a
194
ANDREAS CHAI ET AL.
function of preferences, but a function of rules and knowledge, of which preferences are a subset. The importance of this distinction is that rules and knowledge, unlike the orthodox idea of preferences, are fallible. Thus the basic problem for the consumer in an evolving economic system is the problem of knowing and learning what to want (Earl & Potts, 2004) and then dealing with mistakes along the way. This is not as trivial a problem as it is usually perceived to be in standard micro theory, where consumers want what satisfies their preferences, end of story. From the evolutionary perspective, the ‘what to want’ problem involves more than knowing (or learning) what one’s preferences are; it also entails the strategic coordination of one’s wants with those of other agents, and of not making mistakes in doing so – much in the same way that, as Richardson (1960/1990) emphasizes, firms need to coordinate their investment decisions. Showing up at a cocktail party in an identical dress to someone else can be every bit as embarrassing to two women as the simultaneous proliferation of major investments in capacity to produce a particular good or service can be to the firms who have made the investments, oblivious of each other’s plan. In both contexts, there is more to the decisions than being alert to an opportunity; one must also be able to gauge the likelihood of others being alert to it and be able to act on their alertness. Knowledge influences both consumers’ lifestyle choices and producers’ production decisions. From continually facing new situations, agents, firms and societies learn, and hence their knowledge base continually changes, which in turn changes the way they act, produce, consume and organize in the future. Modern evolutionary economics is thus especially focused on studying how new knowledge affects agents and the system within which they act (Dopfer et al., 2004). In its history, there has been a conscious effort to build an abstract model that can rigorously identify the path through which new elements of knowledge are discovered, selected and adopted by agents, firms and institutions (Dosi, 1982; Nelson & Winter, 1982). The evolutionary theory of consumption focuses on how the demand side evolves. If we understand consumers in a continuous process of ‘learning to consume’ (Witt, 2001), then an evolutionary theory of consumption must seek to incorporate: (1) the selection of problems in respect of which knowledge production (learning) occurs; (2) the behaviour of learning and the complementarities between opportunities; and (3) the management of consumer capital in the face of a turbulent and changing consumption environment to which the consumer’s capital is utilized. A start to applying this alternative view of consumer behaviour was made by Earl (1986), with his conception of a consumer lifestyle. A consumer
Fashion, Growth and Welfare: An Evolutionary Approach
195
lifestyle refers to the way in which the commodities that are inputs into a coherent consumption set (a lifestyle) are connected together as ‘viscous collections of procedures for dealing with fluid situations in which ambiguity is the order of the day’ (Earl, 1986, p. 4). Accordingly, consumption behaviour involves more than simply working out alternative choices and picking the optimal option; it involves navigating a web of everchanging complementarity (cf. Karni & Schmeidler, 1990). Consumption choices are a function of many complex forces, such as social identity in an evolving and turbulent world filled with ever-present uncertainty and the live possibility of costly mistakes: ‘If opportunities are not to be thrown needlessly away, the consumer must be a skilled speculator and strategist’ (Earl, 1986, p. 1). Given this need for strategy, rational consumers structure their consumption behaviour around a set of priorities and goals. Such strategic behaviour helps the consumer to incorporate surprise and anticipate the unexpected, as well as to cope with the inevitable interdependencies that exist among choices. The set of rules the consumer uses to coordinate complementarities is his or her lifestyle (the analogue of productive competence in the theory of the firm). Connections between durable goods are specific structures of complementarity, into which new goods may or may not fit. Thus, consumption sets can be modular, in that one consumption activity cannot simply be substituted for another, but instead the activity may be embedded within a greater consumption strategy that adds up to a consumer lifestyle, as a coherent pattern of connected activities and consumption goods. Insofar as consumers’ lifestyles are based upon decision rules that involve checklists or priority systems to determine whether products fit into their systems, the lifestyle notion leads naturally to specifications in terms of lexicographic orderings instead of utility functions (see Earl, 1986). It is important to note that with this focus on the evolution of knowledge, the primary economic problem no longer concerns consumers spending a constrained amount of income on a range of commodities. Rather, it is one of consumers spending a constrained amount of attention on a range of things that offer to change the consumer’s knowledge base in one way or another. Even with an unlimited amount of wealth to spend on an unlimited amount of goods, economic agents would still face the opportunity cost problem of deciding what pleasures to pursue and for how long (Linder, 1970; Steedman, 2001). The opportunity cost of consumption, and the significance of novelty in dealing with it, would be clear to anyone with a substantial collection of recorded music: each new item stands as a barrier to the consumption of existing items in the collection, so the former will need
196
ANDREAS CHAI ET AL.
to offer more novelty than remains to be gleaned from the latter by repeated consumption. A closely related idea is the concept of bounded rationality, as first conceived by Herbert Simon, which basically states that because agents have a limited amount of reasoning power, decisions incur ‘energy cost’ (Loasby, 2001). If it is cognitively costly to make accurate decisions, then any conception of the consumer perfectly optimizing decisions would require an infinite amount of time and energy. This simply reflects the extension of the fundamental law of scarcity to thought processes. It is impossible knowingly to make an optimal decision in a changing environment, and if particular consumption decisions are perceived to be less important, then less thought will be given to them. Life is a succession of disrupted states of consciousness in which the apparent importance of problems that come to our attention induces corresponding amounts of effort towards solving them or, if they represent a source of cognitive dissonance with drastic implications and little prospect of resolution, justifying turning one’s gaze elsewhere or denying that they exist (Earl & Wicklund, 1999). From this perspective, it is perfectly reasonable to argue that rational economic agents in an evolving economic system will sometimes make mistakes that result in actual utility derived being less than expected utility.4 Making mistakes is inherent in fashion consumption as portrayed in the writings of both Veblen and Scitovsky. For Veblen, purchasing fashionable goods is as much for the satisfaction of other people’s preferences as for one’s own, since the utility one derives from fashion consumption depends on the approval of others. This is an inherent set up for making mistakes if ever there was one. For Scitovsky, fashion consumption is inherently risky, since one can never tell a priori how long an item will provide personal stimulation. If one did have perfect knowledge of a good, it could by definition no longer provide stimulation since, essentially, the demand for novelty is a demand for the unknown. In other words, consumer choice in a complex evolving world is fraught with difficulty. Stability and coherence consists of making connections between how goods fit together, both with each other, and with the social and cultural context of consumption. Mistakes will inevitably be made even by the most rational of consumers, so it becomes important for consumers to develop ways of ensuring that the mistakes that they make do not unduly sour their experience of life in the long term and, above all, are not personally catastrophic. The problem, then, with the old-school literature on fashion is that it was essentially drawn with respect to a static economic background (intermittent not continuous novelty) and without due consequence to the difficulties of
Fashion, Growth and Welfare: An Evolutionary Approach
197
and scope for error when choosing goods and services about which one has little experience and therefore bounded rationality. A boundedly rational agent in a consumption environment that is constantly changing will inevitably make mistakes. Without some mechanism to periodically liquidate these mistakes, a consumer lifestyle will begin to degrade in its social capital until it reaches some threshold of social dysfunction that so passes it, along with its agent, into the realm of being unfashionable and no longer socially observed. That might suit the agent just fine, or it might not: de gustibus. The broader point is to consider the effect this has on both the course corrections of a consumer lifestyle and on the uptake of new technologies into the economy and therefore the sources of economic growth. In an open evolving market economy, there are always new goods and services contesting the markets, so the consumption possibility set is constantly changing, via both entry and exit. This presents the consumer not just with a series of marginal choices (e.g. a new breakfast cereal), but also the possibilities of more radical change in systems of consumption possibilities (i.e. components of a lifestyle) or, even, the chance to make what is for them a revolutionary switch of lifestyle (e.g. move from the city to live by the beach and work remotely via the Internet). But marginal changes to substitute one good for another are always easier than changes to blocks of connected choices. The risk is that new goods and services may only make sense when adopted as a bundle and may never be able to penetrate certain locked-in consumption patterns. In an evolving economy, something must induce agents to revise and update their consumption sets in ways that still leave room for them to learn. Without such an institution, consumers are subject of overspecialization, and their consumption strategies become increasingly inflexible. They risk ending up rather like people with an obsessive-compulsive disorder (Earl, 1986, pp. 164–166) and, as often happens with elderly consumers, becoming detached from modern life, increasingly fearful of venturing out into the world and trying anything new. One way in which markets might promote dynamically efficient behaviour by consumers is if they incorporate some kind of mechanism that forces consumers periodically to re-orientate their learning processes onto different fields (as a kind of positive externality). Schumpeter first suggested this was achieved by the existence of entrepreneurs: ‘new commodities or new qualities or new quantities of commodities are forced upon the public by initiative of entrepreneurs is a fact of common experience’ (Schumpeter, 1928, p. 379). Whilst followers of Schumpeter long have recognized the central function of new goods as a way of evolving the economy, few have really questioned
198
ANDREAS CHAI ET AL.
where the demand for novelty originates, and whether its strength changes over time. Intuitively, for the individual consumer, the demand for novelty is a hazardous want. Not only is there the potential for making mistakes as mentioned earlier, but transaction costs are involved in reconfiguring consumption strategies, and there is far more certainty in ‘sticking to what one knows’. These costs may derive from a number of sources; anything that hinders deviation and promotes conformity is accountable in this matter. Also nothing is said about when the time is right for people to adopt innovations rather than stick to their old strategies. The demand for novelty amongst consumers acts as the essential enabling force that allows innovating entrepreneurs, as creators of novelty, to be successful. In order for consumers to invest in the construction of consumption capital that is relevant to their consumption environments, something must exist to regulate these forces.
4. AN EVOLUTIONARY THEORY OF FASHION AND WELFARE What we argue here is that the introduction of a novel fashion trend into the agent’s environment acts as a potential trigger for consumers to revaluate their consumption strategies in the face of this novel stimulus. The consumption strategies that they adopt in response to this novel stimulus turn into new habits. As the fashion trend becomes more normalized, novelty dissipates. Eventually a point is attained where novelty has dissipated to the extent where a newer stimulus is comparatively novel enough and hence attention arresting for it to be able to force consumers again to re-orientate their strategies. If existing strategies of some consumers are still relatively novel to the extent that it is not worth the transaction cost of recalibrating their strategies to new the stimulus, then novelty will not be adopted by them. It is this decisive occupation of the consumer’s existing attention resources that distinguishes between whether a novel stimulus is a potential or an actualized trigger for lifestyle restructuring. From this framework, we can also understand that the degree of lifestyle complexity is linked to the rate at which the lifestyle is updated. If consumers choose a lifestyle which presents them with a relatively large number of problems, then over time they will have much less attention to dedicate to novel solutions to any particular problem than is available to those who choose much simpler lifestyles. Given this, we would expect much more herd-like behaviour to be displayed by those with complex lifestyles. Such
Fashion, Growth and Welfare: An Evolutionary Approach
199
consumers may have only small areas in which they can develop the expertise to choose for themselves. In principle, they may serve in those areas as trendsetters for their peers, whilst following the latter in other respects. However, in such situations there would be a problem of the overall coordination of the fashionable fit of different but complementary elements of the evolving lifestyle, particularly if specialists in some consumption areas differed in the signals they presented to their peers about ‘the way to go’. Therefore, in practice in such cases, the busy consumers might be expected to delegate to an outside authority with professional expertise the task of ruling on what fitted together. Complex, busy lifestyles seem incompatible with long fashion cycles because of the rapid convergence of behaviour via the use of externally supplied decision rules. By contrast, ‘classic’ styles of consumption that only evolve slowly and are commonly thought of as refined, would seem to be the prerogative of those whose wealth has given them a longstanding ability to consume at leisure. Such ‘old money’ consumers are rich enough to keep many problems at bay – often by following long established social rules – and, having not just ‘arrived’ (unlike the nouveaux riches), they have built up the experience to know how to choose in those areas where the absence of rigid social codes gives them that freedom. Their lack of experience outside their narrow range of deep expertise imparts a profoundly conservative bias to their choices, and their connoisseurship is such that relatively small changes in products that make up their lifestyles will be sufficient to attract their attention. Anything with a particular category that is wildly different from their view of the norm will not capture their attention as it will not fit their classificatory pigeonholes (Hayek, 1952). Fashion cycles and the consumer’s taste for novelty appear to play a very important – not wasteful – role in encouraging flexibility and experimentation in consumer strategies and thereby promoting the development of consumer knowledge and experience. Competition for social standing is not based merely upon displays of how much money one can afford to burn on a particular kind of consumption but also on the ability to display skill in placing the right kinds of fashion bets and not ending up as a ‘fashion victim’ by failing to select neither a fashion rule that is also selected by the vast majority around the same time, nor a strategy whose minority status is regarded as a sign of one being ‘hip’ in Holbrook’s (1995, pp. 319–362) sense of displaying expertise and insight that is ahead of the field. As the impact of the television series Sex in the City on women’s fashion has demonstrated, such rules may embrace both the set of products to purchase and rules for the combinations in which products are consumed.
200
ANDREAS CHAI ET AL.
Seen thus, the fashion-leading consumer has many of the capabilities of an entrepreneur who is alert to gaps in markets and to new opportunities for constructing connections (cf. Earl, 2003). This point has an important implication: the consumer–producer duality that has been a feature of much economics since John Stuart Mill separated demand and supply may be misleading, for entrepreneurs are also consumers. If consumers were predominantly lethargic dullards, growth would be limited not merely due to a lack of interest in new things but also due to a lack of new products in which to be interested. As an anonymous referee for this paper pointed out, ‘Creative and energetic people will express those qualities on both sides of the market, for those two sides are analytical constructions that pertain to the same people’. This implies a new slant on Scitovsky’s The Joyless Economy: those societies that focus on comfort rather than risk-taking pleasure should be expected to have relatively low growth rates due to low levels of entrepreneurial creativity. However, caution is needed here, for comfort-seeking societies may be much more receptive to products developed at home, than to boldly innovative products from more pleasure-focused economies and patterns of trade would therefore not necessarily favour the latter. The former societies could still have fashion cycles, but they would tend to be rather routine, perhaps exemplified by minor but visible tinkering with car designs to signify a new model year. When a fashion cycle comes to an end, those who placed unfortunate bets are put back on a more nearly equal footing with those who succeeded in avoiding being seen, that time around, as fashion victims. For example, if all trousers fall out of fashion in favour of skirts, then it no longer matters that one chose the ‘wrong trousers’ when they were in fashion; the issue now is whether one can make a competent choice of skirt in the eyes of one’s reference group. To be fashionable now, both fashion victor and fashion victim must incur the costs of tooling up for the novel fashion mode. Fashion cycles also play a major redistributing role in society that mitigates their seemingly wasteful ‘throw away’ aspect. The accelerated depreciation of fashion goods enables them to be enjoyed secondhand by consumer subcultures whose members could not hope to purchase them if their early rates of monetary depreciation accurately reflected their physical depreciation.
5. THE IMPACT OF FASHION CYCLES ON AGGREGATE DEMAND AND CREATIVITY The demand-triggering role of changes in fashion mitigates a problem for affluent economies that was recognized by Fisher (1935) and Reddaway
Fashion, Growth and Welfare: An Evolutionary Approach
201
(1937): there is a clash between progress and security when rising incomes open up new consumption possibilities and, with them, new risks. Consumption becomes much more like business investment when unfamiliar new (or previously unaffordable) products are involved, as it does when it has a social dimension and reactions of onlookers cannot be taken for granted. The key problem for sustaining economic growth in affluent economies is that if income is available for discretionary spending, by definition it does not have to be spent. As economic psychologist George Katona (1960) emphasized, the affluent consumer may have the ability to spend (and modern consumers have much more access to credit than they did when Katona was writing), but demand may dry up if they lack the will to spend. Clearly, concerns about the security of income streams that are necessary to service debt or save up for retirement may make consumers unwilling to spend, but so too may seemingly overwhelming tasks of choosing the right product in a functional or social sense. If one does not have to buy something, the problem of choosing what to buy can simply be left in the ‘too hard basket’ for now, but in leaving it there, one is taking away someone else’s income flow and inducement to invest. In terms of our analysis, the forces of fashion can intervene, via the pressure of social competition, to overcome such weakness of will and help to keep the economy closer to its potential growth trajectory.5 At the micro level, fashion cycles are inherently disruptive and attempts by firms to insure against them by diversification carry costs in terms of foregone economies of scope (Kay, 1997). However, at this level, there may be lessons to be learnt from research on the effects of investment spikes at the macroeconomic level, such as those associated with tendencies of firms to invest in plant investments in bursts rather than in small adjustments of capital stock as the neoclassical theory suggests. The theme of lumpy replacement cycles has become increasingly popular amongst macroeconomists who face the task of explaining volatile investment patterns amongst firms (Cooley, Greenwood, & Mehmet, 1997). From this approach, a number of conjectures have been made about a link between macroeconomic fluctuations and investment spikes. Cooper and Haltiwanger first proposed that times of economic downturn are the best times to replace capital stock. They contend that ‘Machine replacement is most likely to occur during downturns where the resource cost replacement is lower (due to low demand and/or high value of leisure) and just prior to upturns where the benefits of replacements are higher’ (Cooper & Haltiwanger, 1990, p. 34, 1993). Indeed, there have been many documented cases where recessions caused by weakening demand, have liquidated all but the most technologically advanced firms. Similarly, Caballero and Hammour
202
ANDREAS CHAI ET AL.
(1994) observe that job destruction is much more responsive than job creation to business cycles, which leads them to argue that recessions are a time of ‘cleansing’ when outdated or unprofitable techniques and products are pruned out of the system.6 Potts (2004) has argued that we might view asset price bubbles in the same way. An asset bubble is a further variant on the liquidationist thesis, arguing that the low cost of finance and the tolerance for mistakes that accrues during a bubble increases the rate of novelty generation and diffusion. This surge of liquidity and experimentation fuels investor (and consumer) demand, at the same time promotes the restructuring of the economic system. The micro-level counterpart of this line of thinking involves applying Burton Klein’s (1977) analysis of the link between product lifecycles and economic growth to the case of fashion. Klein’s key theme is that uncertainty about what will become the dominant standard for a product genre is good for productivity growth, and that productivity growth tends to slow down once a standard has emerged and production has become concentrated in the hands of the firms who survived because they worked out the least-cost way of making products in terms of this standard. The crucial driving force of innovation are the rich pickings that await those who place their bets on the winning standard and work out how to make it win by solving technological problems. Once the big challenges have been dealt with and it is clear that the winning standard of, say, the motor car involves internal combustion engines rather than steam or electric power, steering wheels rather than tillers, and so on, Klein suggests that the survivors will recognize that they are each big enough to get the capabilities and funds to replicate any major changes their rivals venture. Competition then becomes much less technologically aggressive and changes tend to be more marginal rather than revolutionary in nature. Major changes will thus tend to be made only if originated by outsiders. If applied to the context of fashion cycles, Klein’s argument must be cast more in terms of uncertainty about aesthetic issues than technological ones. But in this context it involves both producers and consumers. When a product goes out of fashion, its producers and past consumers are forced into problem-solving mode. Thus although the liquidation of some of their assets is costly to them, the process of dealing with this setback is likely to result in them emerging with enhanced knowledge. Furthermore, since the end of one fashion cycle is not the beginning of a clearly defined new fashion but a time for experimentation within a new genre, there is everything to play for. It therefore pays to be bold and creative, whether as a consumer or as a producer, given the possibility of becoming a leader. Even those who
Fashion, Growth and Welfare: An Evolutionary Approach
203
lose their jobs in firms whose products have fallen out of fashion without being replaced by something else that is successful may emerge from the experience wiser and more prosperous and/or content: to find a new job may involve relocation and retraining and generally getting out of the rut they were in. The revolutionary change in the popular music market that took place in the second half of the 1970s is an excellent example of this process. It involved the displacement of progressive rock, based on performer virtuosity and elaborate stage design, by the ‘new wave’ of much simpler music such as punk rock. Suddenly, major record companies found their established acts were not selling and instead consumers were buying music by new performers that was being released by small, independent record companies. This was also a time of great opportunities for performers with no track records in the music industry. Record companies and consumers alike thus faced an explosion of new acts offering creative new sounds and images, but it was far from clear to either group where they should invest if they were to avoid financial and/or social embarrassment: would it be wise to invest money in the output of a band such as The Sex Pistols whose members were trying to appeal by breaking all the established rules, or would it better to invest in music that was less pompous and arty than that of bands such as Yes or Emerson, Lake and Palmer but was at least being produced by musicians who had some skill in playing their instruments and writing well-crafted, catchy songs? Further uncertainty was caused by technological change that resulted in bands using cheap synthesizer keyboards starting to challenge those based on guitars. The upheaval also forced the former progressive rock giants to experiment and reinvent themselves, with the result that some bounced back several years later with huge success and a wider audience than before – without ‘The Sex Pistols’ ‘Anarchy in the UK’, Yes might never have come up with ‘Owner of a Lonely Heart’. And in the midst of all this, the major record companies learnt a lot about just how lax they had been in containing recording costs whilst progressive rock epics were being hatched.
6. CONCLUSION Where the standard view of fashion in microeconomics and consumer theory views fashion as a profligate indulgence, we have argued that it plays a more positive role in stirring consumers into actions from which their pools of knowledge and range of experience may grow. Just as an increase in the
204
ANDREAS CHAI ET AL.
strength of competition may prompt decision-makers to explore ways of increasing productivity, so a change in relative competitive strength between status-conscious consumers may force them to rethink their choices. Just as firms in featherbedded markets may fail to develop new knowledge, so consumers who opt out of social competition and take the ‘quiet life’ may fail to develop their ranges of experience and capabilities. Such a lifestyle may appeal to older consumers: they have established who they are in the social order and, with fewer years of life remaining, they have less of an incentive to take risks associated with experimentation to acquire new capabilities as consumers. Not so the young, for whom there are higher payoffs to being ‘hip’ and acquiring a reputation of being ahead of the pack or, at least, to know what is fashionable. Mistakes are inevitable in the process of social competition. Sometimes we buy consumer goods that just don’t fit into our lifestyles, things that just don’t connect or enable us to connect socially: the wrong trousers or the wrong lounge-suite or cell-phone or club membership or car and indeed any durable good in some measure. Development in consumer lifestyles is a process of re-coordination of a complex system of consumer durables. This growth process is facilitated by periodic liquidation for exactly the same reason that macroeconomic growth is also facilitated by periodic liquidation, namely that it lowers the overall cost of transformation. Fashion is a mechanism that is a part of this process and so fashion cycles are necessary components of macroeconomic growth. Fashion not only enforces flexibility in consumer lifestyles but also has a positive distributional effect on consumer welfare by erasing past consumer mistakes as well as minimizing the opportunity cost of adopting novelty. Continuous economic growth requires consumers to have a continuous will to buy, learn and take risks. Risk-taking behaviour inevitably causes mistakes, which are a necessary byproduct of economic growth. What is needed, therefore, is a mechanism to erase consumer mistakes in order to regenerate the incentive for them to continue learning. The modern social phenomena of fashion enables economic growth by providing consumers the twin incentive of both abandoning old fashion rules and adopting new rules through (a) periodically liquidating dated fashion goods and their related mistakes and (b) providing alternative goods that, thanks to standardization, cater for the varying risk preferences of consumers. Fashion trends can thus be understood as learning trajectories by re-orientating consumer attention into new areas of learning. Through the working of social pressure, they periodically provide a fresh and self-regulated impetus for consumer learning. Fashion cycles periodically loosen the constraints
Fashion, Growth and Welfare: An Evolutionary Approach
205
that accumulate on the demand side and thereby facilitate the process of economic growth and personal development.
NOTES 1. See Saviotti (1996), Bowles (1998), Witt (2001), Potts (2000), Earl and Potts (2004). 2. Improved functionality can entail both better performance on familiar dimensions, as emphasized by Lancaster (1966) or features never before offered: see Bianchi (2002). 3. The liquidationist thesis, first advanced by Friedrich Hayek, Joseph Schumpeter and Lionel Robbins, holds that far from being entirely negative in consequence, recessions are actually beneficial to the economy in that the low real prices of factors and resistance to change make them effective periods of ‘structural cleansing’ in a macroeconomy. 4. Note that mistakes may also be positive, i.e. accidentally acquiring more utility than anticipated, but these do not really pose problems for agents and the economy. 5. Note that we are not saying that the forces of fashion assure maximum possible growth and full employment. The case of Japan is a reminder that an economy can be very focused on fashion and novelty and yet still have high rates of saving and suffer from economic stagnation, but its stagnation would have been far worse without fashion as a driving force, both in terms of domestic demand and in terms of the benefits derived from the export of innovative products. 6. In a later study of US manufacturing data, Caballero and Hammour (1999) found evidence for a ‘reverse-liquidationist’ position, which held that recessions can be associated with a ‘chill’ in the restructuring process, rather than increased ’turbulence’. Thus an increased period of liquidation does not necessarily lead to a increased period of restructuring. How well these two are connected will depend on the institutional environment in which the firms operate.
ACKNOWLEDGMENT We are grateful to Brian Loasby and to an anonymous referee for extensive and very helpful comments on an earlier version of this paper. However, the usual disclaimer applies.
REFERENCES Bianchi, M. (2002). Novelty, preferences, and fashion: When goods are unsettling. Journal of Economic Behavior & Organization, 47, 1–18.
206
ANDREAS CHAI ET AL.
Bowles, S. (1998). Endogenous preferences: The cultural consequences of markets and other economic institutions. Journal of Economic Literature, 36, 75–111. Caballero, R., & Hammour, M. (1994). The cleansing effects of recessions. American Economic Review, 84, 1350–1368. Caballero, R., & Hammour, M. (1999). The cost of recessions revisited: A reverse-liquidationist view. National Bureau of Economic Research, Working Paper, 00-02. Cooley, T., Greenwood, J., & Mehmet, Y. (1997). The replacement problem. Journal of Monetary Economics, 40, 457–499. Cooper, R., & Haltiwanger, J. (1990). Inventories and the propagation of sectoral shocks. American Economic Review, 80, 170–190. Cooper, R., & Haltiwanger, J. (1993). The aggregate implications of machine replacement: Theory and evidence. American Economic Review, 83, 360–382. Dopfer, K., Foster, J., & Potts, J. (2004). Micro – meso – macro. Journal of Evolutionary Economics, 14, 263–280. Dosi, G. (1982). Technological paradigms and technological trajectories: A suggested interpretation of the determinants and directions of technical change. Research Policy, 12, 147–162. Douglas, M., & Isherwood, B. (1978). The world of goods: Towards an anthropology of consumption. New York: Basic Books. Earl, P. E. (1986). Lifestyle economics: Consumer behaviour in a turbulent world. Brighton: Wheatsheaf. Earl, P.E. (2003). The entrepreneur as a constructor of connections. In: R. Koppl (Ed.), Austrian economics and entrepreneurial studies: Advances in Austrian economics (Vol. 6, pp. 113–130). Oxford: JAI/Elsevier. Earl, P. E., & Potts, J. (2004). The market for preferences. Cambridge Journal of Economics, 28(4), 619–633. Earl, P. E., & Wicklund, R. A. (1999). Cognitive dissonance. In: P. E. Earl & S. Kemp (Eds), The Elgar companion to consumer research and economic psychology (pp. 81–88). Cheltenham: Edward Elgar. Fisher, A. G. B. (1935). The clash of progress and security. London: Macmillan. Foley, C. (1893). Fashion. The Economic Journal, 3, 458–474. Hayek, F. A. von (1952). The sensory order: An inquiry into the foundations of theoretical psychology. London: Routledge. Holbrook, M. B. (1995). Consumer research. Thousand Oaks, CA: Sage. Karni, E., & Schmeidler, D. (1990). Fixed preferences and changing tastes. American Economic Review, 80, 262–267. Katona, G. A. (1960). The powerful consumer. New York: McGraw-Hill. Kay, N. M. (1997). Pattern in corporate evolution. Oxford: Oxford University Press. Klein, B. H. (1977). Dynamic economics. Cambridge, MA: Harvard University Press. Lancaster, K. (1966). Change and innovation in the technology of consumption. American Economic Review, 56, 14–23. Linder, S. B. (1970). The harried leisure class. New York: Columbia University Press. Loasby, B. (1999). Knowledge, institutions, and evolution in economics. London: Routledge. Loasby, B. (2001). Cognition, imagination and institutions in demand creation. Journal of Evolutionary Economics, 11, 7–21. Nelson, P. (1970). Information and consumer behavior. Journal of Political Economy, 78, 311–329.
Fashion, Growth and Welfare: An Evolutionary Approach
207
Nelson, R., & Winter, S. (1982). An evolutionary theory of economic change. Cambridge, MA: Belknap Press of Harvard University Press. Potts, J. (2000). The new evolutionary microeconomics. Cheltenham: Edward Elgar. Potts, J. (2004). Liberty bubbles. Policy (CIS), 20, 15–21. Reddaway, W. B. (1937). Special obstacles to full employment in a wealthy economy. Economic Journal, 47(June), 297–307. Richardson, G. B. (1960). Information and investment. Oxford: Oxford University Press (republished, 1990). Saviotti, P. (1996). Technological evolution, variety and the economy. Cheltenham: Edward Elgar. Schumpeter, J. A. (1928). The instability of capitalism. Economic Journal, 38, 361–386. Scitovsky, T. (1976). The joyless economy: An inquiry into human satisfaction and consumer dissatisfaction. Oxford: Oxford University Press. Steedman, I. (2001). Consumption takes time. London: Routledge. Veblen, T. (1899). The theory of the leisure class. New York: Macmillan (London: Penguin, 1994). Witt, U. (2001). Learning to consume: A theory of wants and the growth of demand. Journal of Evolutionary Economics, 11, 23–36. Witt, U. (Ed.) (2003). The evolving economy: Essays on the evolutionary approach to economics. Cheltenham: Edward Elgar.
This page intentionally left blank
FASHION: WHY PEOPLE LIKE IT AND THEORISTS DO NOT Luciano Andreozzi and Marina Bianchi ABSTRACT One of the many paradoxes of fashions is that consumers’ choices change rapidly and with an astonishing degree of synchronization. What is successful or socially acceptable in one period is considered the opposite in the next. This paradox has brought economists and other social scientists to conceive of fashions and fads as one of many forms of irrational behavior. Herd behavior and weakness of will, a desire to conform or, conversely, to distinguish oneself, have all been invoked to explain the rapid evolution of modes of behavior that emerge and more or less suddenly disappear. In this paper we try to show that fashions, even if fragile and transient, might nonetheless be rational. It is a rationality, however, that has to include something overlooked in most economic writing: the desire for novelty and variety. In fashions this desire takes the form of coordinated behavior that both facilitates consumption and destroys its novel content, thus paving the way for new fashions to appear.
1. INTRODUCTION Though we are used to associate fashion with the realm of luxury goods such as haut-couture dresses and expensive jewelry, fashion cycles can be The Evolution of Consumption: Theories and Practices Advances in Austrian Economics, Volume 10, 209–229 Copyright r 2007 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 1529-2134/doi:10.1016/S1529-2134(07)10009-0
209
210
LUCIANO ANDREOZZI AND MARINA BIANCHI
observed in all sorts of goods and activities. From popular literature and music to entertainment and leisure-time activities, from art and decoration to body shapes and features, we witness trends, genres, and styles that change in succession, sometimes disappearing forever, more often reviving in new guises. Fashion seems to be ubiquitous and a permanent dimension of human behavior.1 Given the pervasiveness of fashion, it comes as no surprise that there is a long tradition of economic speculation about it, although works specifically devoted to fashion represent only a small fraction of the much larger literature on consumer behavior. Within this literature there is a remarkable degree of agreement on what should be considered as a fashion. Fashions are: 1. Short-lived. There is no such a thing as a timeless fashion. Shared patterns of consumption that are long-lived come under the headings of ‘‘traditions’’ and ‘‘usages’’ rather than fashions. Being transient and ephemeral is the most evident characteristic of fashions, almost to the point of defining them. 2. Cyclical. Though short-lived, fashions tend to follow patterns that are cyclical. Fashionable goods and activities disappear, only to be replaced by new ones that will follow a similar pattern of success and decline. In this process, goods that have once been fashionable are often rescued from their past and given new life, albeit in somewhat different form. Fashion cycles include many such revivals. 3. Synchronized. This is fashion’s most striking aspect: changes of consumer behavior, though cyclical and short-lived, still display significant coordination. Songs become popular and ubiquitous for a period and then suddenly disappear, as if by common agreement. The same can be said of hairstyles, popular restaurants, cafe´s and clubs, parlor games and tourist resorts, movies and novels. In all these cases, individual consumers seem prone to switch to a different form of consumption just when everyone else is ready to switch as well.2 It is probably because of this last characteristic that economic models of fashion are usually grounded on ‘‘social’’ preferences, rather than on the individual preferences that dominate in all other models of economic behavior.3 When following fashion, individuals are assumed to seek status and distinction or, conversely, to emulate and conform. Individual preferences are replaced with those of others – though where these others derive their preferences usually remains unexplained. We take a different tack. We present a model of fashion cycles that recovers the idea that individuals, even when following fashions, are motivated by their own preferences and the intrinsic rewards of their choices. At the same
Fashion: Why People Like It and Theorists Do Not
211
time, we assume that individuals interact with each other and coordinate either intentionally – through mutual emulation, innovation, and rivalry – or unintentionally as a cumulative effect of their individual choices. We assume also that choices have a history and show interdependencies between past and present because of learning and habit formation (Becker & Murphy, 1988; Pollak & Wales, 1992; Becker, 1996; Conlisk, 2003). These alternative assumptions liken the study of fashion to the spontaneous emergence of social coordination that has been a popular theme in the institutional economics of the last two decades. However, we argue that what distinguishes fashion from other forms of coordinated behavior such as language, rules of etiquette, and traffic rules is that fashion is novelty-driven. Any new fashion starts with a rupture, a more or less strong break with the past. And this is the reason for its appeal. This approach presents an obvious paradox: if fashion expresses an individual desire for novelty or innovation, why do we witness a synchronization of behaviors which so closely resembles conformism? The answer we suggest is simple. Since novelty cannot be judged and appreciated without a reference point, consumers, especially when engaged in forms of consumption that are more social and visible, inevitably refer to each other in order to compare and understand what is new relative to what has been enjoyed before and by others. In this process, fashions work as points of attraction. They provide that shortcut of experience, thanks to which what is new can be perceived neither as too foreign and therefore unpleasant, nor as already old and boring. However, this balance is not meant to last. The more a fashion diffuses and is repeated over time, the more its novelty erodes and loses its appealing characteristics. That is when a new fashion appears.4 The paper proceeds as follows. Section 2 presents a brief review of the economic literature on fashion and our alternative approach. Section 3 analyses the motivational basis that links fashion to change. Section 4 gives a general and formal representation of our own model of fashion cycles. Section 5 offers some illustrations of fashion trends and cycles and shows how the interplay of novelty and social coordination can explain both. Section 6 concludes.
2. THE ECONOMIC LITERATURE ON FASHION Jon Elster once noticed that ‘‘one of the most persistent cleavages in the social sciences is the opposition between two lines of thought conveniently associated with Adam Smith and Emile Durkheim, between homo oeconomicus
212
LUCIANO ANDREOZZI AND MARINA BIANCHI
and homo sociologicus. Of these, the former is supposed to be guided by instrumental rationality, while the behavior of the latter is dictated by social norms’’ (Elster, 1989, p. 100). The differences between homo oeconomicus and homo sociologicus are best appreciated by noticing that ‘‘social preferences’’, as opposed to ‘‘individual preferences’’, play a much larger role in sociology than in economics. While sociologists are more likely to explain individuals’ behavior in terms of their desire to conform, or to distinguish themselves from other people (Jones, 1984), economic research is mostly driven by the assumption that people have well defined individual preferences and choose accordingly. When it comes to fashion, however, this divide blurs because economists themselves tend to appeal to explanations that have a sociological flavor. Most of economic theorizing about fashion is inspired by heterodox authors such as Thorstein Veblen, who in turn is closer to a sociologist such as Georg Simmel than to standard neoclassical economics which he strongly criticized (Tilman, 1998). In Veblen’s theory, fashion goods are just an instance of ‘‘conspicuous consumption’’, which is the larger set of consumption activities that people display to mark their social distinction. Veblen’s approach is interesting here because, in their endless competition for status, consumers’ personal preferences play little or no role. His consumers are ready to adopt any consumption good, as long as it can be used to signal their social position. As a consequence, individuals may never end up consuming something they actually like. In fact, the very notion of individual preferences becomes redundant when discussing fashionable (or, more generally, luxury) goods. The question is not whether a consumer prefers to wear a black or white suit. The question is: what color is ‘‘in’’ at the moment among the ‘‘right’’ people?5 This view of fashion as invidious competition has been the main inspiration for later theorizing on this topic. Leibenstein (1950), in an influential article, distinguished between a ‘‘bandwagon effect’’, that occurs when people try to conform and imitate the majority of their peers, and a ‘‘snob effect’’, that takes place when people do the opposite. This distinction tells us that the internal dynamics of fashion are driven by the desire of the few to distinguish themselves from the mass, coupled with the desire of the mass to emulate the members of this small group. In a later article in a similar vein, Dwight E. Robinson was thus able to define fashion as a ‘‘race of appearances’’ – an expression he takes from the Romantic writer and essayist William Hazlitt (Robinson, 1961). Things have not changed. The relatively large number of papers on fashion published in the last decade formalizes one aspect or another of Veblen’s original approach. In these models, consumers are usually divided into a
Fashion: Why People Like It and Theorists Do Not
213
lower and an upper class, and it is assumed that people belonging to the lower class try to imitate people belonging to the upper class, while people belonging to the upper class try to distinguish themselves from people belonging to the lower. Karni and Schmeidler (1990), Cowan, Cowan, and Swann (2004), and Corneo and Jeanne (1999) present models along these lines. They show that the interaction between snob and bandwagon effects alone is sufficient to keep the population of consumers in an endless cycle of fashionable consumption. Since differences in the relative prices of goods play no role in this approach, the ‘‘conspicuous consumption’’ aspect of Veblen’s theory is not modeled here.6 There is a strand of the literature, however, that formalizes that notion too, though without any direct reference to fashion. In these models, as in Veblen’s original analysis, individuals are assumed to buy expensive and useless goods in order to display their status, wealth, etc. (see, e.g. Frank, 1985; Congleton, 1989; Bagwell & Bernheim, 1996; Pesendorfer, 1995; Frijters, 1998).7 Compared to traditional models of consumer choice, models of fashion have made an important contribution. They have taken consumers away from their solitary choices and started to explore the modalities of their interaction. At the same time, however, they portray fashion as an endless chain of mutual reactions that are disconnected from any intrinsic utility of choice. There is no intrinsic merit in goods thought to be fashionable unless and until the upper classes and trendsetters declare them to be so. Yet, even casual observation shows that frequently just the opposite is true. In the 1960s, young people listened to the Beatles and Rolling Stones, girls wore miniskirts, and Frisbee and skateboard were among the most popular outdoors games. However, it is hard to imagine that all these activities enjoyed widespread success without their intrinsic characteristics and the individual preferences of their adherents playing any role. Importantly, however, in the above examples fashions began inauspiciously, in unknown pubs or clubs, with designs for young girls unrelated to the seasonal cycle of haut-couture and youths seeking enjoyment in inexpensive sports. Faded blue jeans, espadrilles, sneakers, and t-shirts were all born out of novelty, convenience (in comfort and price), and protest, rather than as down-market imitations of the luxurious.8 Often, indeed, new trends in music, styles of furniture, and dresses are first discovered in the world of self-made music, in flea markets and junk shops, and in sports and outdoor activities. Only later do they become reproduced as fashionable items for those with the wherewithal. The spread of the novel in eighteenth-century Europe, to take a less familiar example, started largely
214
LUCIANO ANDREOZZI AND MARINA BIANCHI
because women unfashionably abandoned their devotional books and joined a much less respectable and frowned-upon world of adventures, facilitated in this process by the rise of traveling booksellers and the availability of cheap lending libraries. What these examples suggest is that consumers are more active and entrepreneurial than usually portrayed in models of fashion. De Vany (2004) has shown this in the case of motion pictures. A film’s ultimate success is not predicted by its budget, stars, advertising campaigns, multiple theater openings or large first-weekend box office receipts. Word of mouth communication plays a much larger role. The fate of a movie at the box office is usually decided only by the fifth or sixth week, when moviegoers have started to share their preferences and reactions; then a film’s success starts to be recognized. Yet what, at that point, might appear to be herding behavior, as in an informational cascade, is, in fact, informed consent and consensus.9
3. FASHION AND NOVELTY What is it in fashion that attracts? If it is neither simply status nor, obviously, only functionality, what is it that motivates fashion? Famous early researchers in the study of the motivational mechanisms underlying aesthetic preferences were Wundt and Fechner, in the second half of the nineteenth century (see Crozier & Chapman, 1984).10 Recent experimental research coming from the field of behavioral economics, experimental psychology, and the neurobiology of the brain, has added relevant and new understanding to the topics that interested those pioneers. Despite differences of approach and theoretical aims, these studies show a rather striking convergence of results when they spell out the motivational variables that underlie an aesthetic experience. Supported by a rather strong body of experimental evidence, they show that positive hedonic values correspond to those variables of the stimulus potential that are linked with change, contrast, and conflict (see Camerer, Lowenstein, & Prelec, 2005, p. 28). Specifically, these variables can be listed as novelty, surprise, variety, complexity, uncertainty, synergies, and mystery (see Berlyne, 1971; Berlyne & Madsen, 1973; Apter, 2001; Barkow, Cosmides and Tooby, 1992; Kaplan, 1992; Scitovsky, 1992 [1976]). Positive hedonic responses, however, do not increase monotonically with increases in the stimulus potential. Both low and high levels of arousal are aversive. Pleasure is maximal for intermediate levels of arousal, when novelty, variety, uncertainty, and so on, are felt to be neither too high nor too low (see
Fashion: Why People Like It and Theorists Do Not
215
Bianchi, 1998, 2003). ‘‘Low’’ and ‘‘high’’ levels however are meaningless without a reference point in respect to which they are valued. Novelty and the other variables of change as well are relative concepts. Specifically, we can distinguish three main factors on which they depend: (1) the time and frequency of exposure; (2) the individual’s accumulated experience and knowledge; and (3) the surrounding environment, such as other people’s experiences and modes of behavior.11 Both the passage of time and the accumulation and diffusion of knowledge shift the reference point within which an activity is assessed. Repetition and acquaintance can cause adaptation and satiety, transforming what was once exciting into something boring. But the reverse can also occur, as when an already experienced activity becomes exciting again because time and knowledge have added as yet unexplored dimensions. Depending on the reference point, then, the same variable can have a double and possibly contrary impact on the affective assessment of an experience. As a result, the same event can be either positively or negatively felt. Exposure to a new dress at the beginning of the season, when its features are new and few people have adopted it, can be very enjoyable. Exposure to the same dress at the end of the season can be off-putting. Fashions provide stimulation for consumption by playing on all three dimensions of novelty. They both familiarize – through coordination and diffusion, through quotations or revivals of old styles and trends – and de-familiarize – through the introduction of new features, and through individual variations and differentiations.12 And at their peak, fashions provide optimal arousal levels (where novelty is neither too high or too low). These optimal levels, however, are necessarily ephemeral because, through continual repetition and diffusion, fashions end up destroying novelty. Our model, though highly stylized, seeks to capture these characteristics and to show that fashion emerges even in the absence of conformism.
4. THE MODEL 4.1. General Description In our model we assume that individuals’ choices are limited to just two alternatives. Think of them as alternative styles in clothes (e.g. fitted or loose, formal or informal), colors in furniture (neutral or colored), types of cooking (Italian vs. Japanese), or children’s names (Ernest vs. Oscar). Utility payoffs depend on the relative novelty of each alternative, and novelty in its turn
216
LUCIANO ANDREOZZI AND MARINA BIANCHI
depends on the frequency with which each alternative has been consumed in the past. We assume that repeating the same strategy over time has a negative effect on utility, because of increasing boredom. (After having worn black for a whole season, one longs for colors, and vice-versa.) The consumer thus faces a scenario in which the strategy that has been chosen, and repeated, slowly loses its novelty-content and, with it, its attractiveness. By contrast, the strategy that has not been chosen somehow ‘‘recharges’’ its novelty content and gains in attractiveness. (After a period of abstinence one rediscovers forgotten tastes, songs, authors, colors.) In choosing between the two available strategies, individuals are not independent of each other, however. If they were – for example if they lived on separate islands – collective phenomena such as fashions would be utterly impossible. But people do not live on separate islands. Many forms of consumption are enjoyed together and in public spaces: at work, in the streets, at entertainments and in other leisure contexts. Moreover, social interaction is just the place for emulation and learning. To abandon the old for the new can be attractive but also source of mistakes and utility losses. Social interplay can help reduce the risk of error. We assume, in fact, that individuals are rational in the traditional economic sense that they choose the alternative they prefer, given their past history of consumption. But they are not completely rational either, because we also assume that occasionally they make mistakes that lead them to select a strategy which is not optimal given their current preferences. These mistakes are more likely when the chosen strategy has lost part of its appeal but the new, just because it is new, is not appealing enough. It is in this situation of near-indifference, where individuals find themselves unable to choose between the two strategies, that they start, with a certain probability, to look at each other in order to learn and decide which to choose. Our simulations show that the emergence of fashion cycles reflects two main variables: (a) the degree of interaction when preferences are not yet strong and (b) the degree to which novelty erodes through repeated consumption. 4.2. Formal Features Time in the model is discrete. For expository convenience time is divided into days, although any other unit of measure would do. So, every day each consumer must choose between two strategies, S1 and S2. Each strategy’s payoff depends upon the frequency with which that strategy has been chosen in the past. To model this we follow the classical research in timeinterdependent choices due to Herrnstein (see Herrnstein & Prelec, 1992),
217
Fashion: Why People Like It and Theorists Do Not
although our model is also compatible with other approaches such as the one pioneered by Becker and Murphy (1988). Current payoffs are thus influenced only by the choices made in the last m rounds (or days). Let m1 and m2 be the number of times the consumer has chosen strategies S1 and S2 respectively in the last m days and let xi ¼ mi/m be the fraction of times strategy i ¼ 1,2 has been chosen. The payoff to each strategy is a decreasing function of the frequency with which that strategy has been played over time. That is, there are two strictly decreasing utility functions ui(xi) that represent the utility obtained from strategy i, when it has been played with a frequency xi over the last m days. It will be mathematically convenient to assume that ui(?) are the same for both strategies and have a very simple linear form, ui(xi) ¼ 1 xi. Our hypotheses are summarized in the figure below. Each alternative yields its maximum payoff when it has never been used in the last m rounds (so that mi/m ¼ xi ¼ 0), and yields its minimum payoff (zero) when it has been played for each of the last m rounds. Again for mathematical convenience, payoffs have been normalized so that the highest is equal to 1 and the lowest is equal to zero, although nothing specific depends upon this choice. 1 0.8
U2m2
U1m1
0.6 0.4 0.2 0.2
0.4
0.6
0.8
m1 1 m1 + m2
An individual’s behavior can be modeled by means of what in the literature is known as the log linear response model, originally proposed by Blume (1995) in the context of evolutionary game theory (see also Young, 1998). At every iteration, each player chooses with probability (1 d) a strategy, which maximizes his utility function. With probability d the consumer picks a strategy at random. This class of models differs from other models of limitedly rational behavior in that it is assumed that the probability d of not choosing the most preferred course of action, i.e. of making a mistake, decreases exponentially with the utility loss due to a non-optimal choice. In other words, mistakes are more likely when the utility losses are small, and become increasingly likely as losses approach zero. In terms of the figure above,
218
LUCIANO ANDREOZZI AND MARINA BIANCHI
this last happens in proximity to the intersection between the two lines, that is when both strategies have been chosen approximately half of the time in the last m rounds. To put this formally, define D ¼ 9U1(m1) U2(m2)9 D is the loss the consumer suffers when she plays a non-optimal strategy. We assume that d ¼ eD, for a fixed e>0. As a consequence, when D is large, consumers are likely to choose their most preferred course of action. In our specification of the model, the largest difference in utility is 1, which happens when one of the two actions has always been used in the last m repetitions and the other never (in the figure above this happens when mi/m is close either to zero or one). In this case, the probability of making a choice at random is just e. On the other hand, when consumers are indifferent between the two alternatives, that is when x1 ¼ x2 ¼ 1/2 and D ¼ 0, their choices are entirely random, because d ¼ e0 ¼ 1. Standard log linear response models assume that random choices are uniformly distributed among the alternatives. We depart from this approach by assuming that, when deviating from the optimal choice, agents are partly influenced by the individuals they interact with. We capture the influence of other people’s behavior in the following way. When choosing a strategy at random (which happens with probability d) each agent has a probability l of picking the same strategy another individual taken randomly from his population is using. l then measures the degree of social interaction. Two points are worth stressing. First, our way of introducing social interaction does not require any fashion leader (though it does not exclude it). Each individual has equal impact and may both follow and be followed. The traditional distinction between upper and lower classes (or followers and trendsetters) plays no role here. Second, we assume that individuals choose mostly according to their own current preferences. It is only when facing near indifference that suboptimal choices become relevant, and it is only in these cases that individuals’ behavior can be influenced by the choices of other people. When preferences are clearly defined, individuals’ behavior is basically driven by themselves.
4.3. Results We run extended simulations of this model by varying the two parameters: m and l. m represents the number of repetitions it takes to decrease from its
Fashion: Why People Like It and Theorists Do Not
219
maximum value to zero the utility payoffs associated with a certain strategy. When m is small, the marginal impact of each choice on the satisfaction a consumer derives from each alternative is large. For example, m ¼ 2 implies that it takes only two consecutive choices of one alternative to drive to zero the utility a consumer gets from choosing it. This is the case of goods and experiences that tire very quickly because they are too often repeated or are too simple. A larger m represents the opposite situation in which it takes a large number of consecutive choices of the same alternative to reduce its utility to zero. A varied diet, a versatile dress that mixes easily and has multiple uses, a game with hidden surprises, a research project that is rich in developments, are all examples of experiences that engage and stimulate longer when repeated.13 Parameter l determines the degree to which consumers interact when facing the possibility of choosing the wrong strategy. For goods in which the social element of consumption is negligible,l will be small. Hammers, scissors, and common domestic appliances of any form belong to this class of goods. l will tend to be large for goods that are frequently consumed in groups (such as sports and film showings), goods that are visible, so that people consume them by simply interacting with other people (e.g. dresses worn in public, music experienced at a concert), and so on. Of course, when l ¼ 0 there is no interdependence among individuals. In this case we would expect no coordination to emerge, because each individual will choose of his own accord. We expect some form of coordination to emerge for larger values of l. All our simulations assume a population of 10 identical individuals facing the same value of m and l and with a fixed e ¼ 0.05. A very low value of e insures that most of the choices are made according to individual preferences, since the fraction of random choice will typically be low. In the presence of clearly defined preferences (that is when D approaches one) only 5 choices out of 100 will be random. The figure below represents a run for m ¼ 10 and l ¼ 0.4. On the x-axis t represents time, on the y-axis n1 represents the number of individuals within the population who adopt strategy S1. One can easily see that individuals’ behaviors fail to coordinate on a clear pattern or fashion. While individuals continue to oscillate between the two alternatives, the oscillations do not synchronize and they do not appear at the aggregate or social level. This is hardly surprising since both e and l are relatively small. Of the few choices that are random, only 40 percent are driven by mutual imitation.
220
LUCIANO ANDREOZZI AND MARINA BIANCHI n1 10
ε = 0.05; λ = 0.4; m = 10
8 6 4 2 t 50
100
150
200
The next figure represents a situation in which all parameters are unchanged, but the probability of imitation is larger, l ¼ 1. Now individual’s choices are coordinated in a well-defined pattern: the entire population switches regularly between the two extremes in which all individuals play the same strategy simultaneously. n1
ε = 0.05; λ = 1.; m = 10
10 8 6 4 2 t 50
100
150
200
Finally, the two pictures below represent the same set of parameters, but with larger values of m. In the first picture m is equal to 14, and in the second to 20. When m ¼ 14, after a short transitory state in which no clear cycles appear, the population settles into a relatively regular cycle of fashion. The thing to notice here is that each cycle lasts longer now than when m was equal to 10. To see this, consider for example the second half of the simulation, from round 100 onward. When m ¼ 10 at least 8 cycles can be counted. When m ¼ 14, the number of cycles is reduced to six. With a larger m there seems to be a smaller number of cycles, each of which has longer duration.
221
Fashion: Why People Like It and Theorists Do Not n1 10
ε = 0.05; λ = 1.; m = 14
8 6 4 2 t 50
100
150
200
The picture below represents a still larger value of m ¼ 20. It shows a new pattern. After a brief period in which the fashion cycle emerges and the population gets locked into clearly defined oscillations, the pattern breaks and reemerges later. Overall, there is less coordination of individuals’ choices for this value of m than with smaller ones. n1 10
ε = 0.05; λ = 1.; m = 20
8 6 4 2 t 50
100
150
200
Two lessons can be learned from these preliminary results. First, fashion is more likely to emerge when people are prone to influence each other (i.e. when l is large) and it is less likely to emerge when past choices have only a small impact on current utilities, that is when m is large. Furthermore, fashion cycles (when they exist) tend to last longer when m is large than when m is small. To explore these two effects somewhat more systematically, let us first define when a good is fashionable. We say that a good is fashionable when more than 90 percent of the population adopts it. We have run simulations for values from 2 to 20 for m, and from 0.8 to 1 for l. For each simulation we computed the first 200 rounds.
222
LUCIANO ANDREOZZI AND MARINA BIANCHI
The results of the average of 100 trials of these simulations is shown below. The vertical axis (labeled ‘‘Fashion’’) represents the fraction of total rounds in which one of the two goods has been fashionable. The picture shows that coefficient l has a fairly predictable impact on fashion: for any value of m, a larger value of l increases the number of rounds in which more than 90 percent of the population chooses simultaneously the same alternative. In other words, and other things being equal, the periods in which one of the alternatives is fashionable are more frequent the larger is the value of l. The impact of m is more complex. For very low levels of m, fashion cycles will not emerge even if l is close to one. Fashion becomes dominant for values of m around 10 but then declines somewhat. Intuitively, for extremely low values of m, individuals end up having always clearly defined preferences for doing the opposite of what they have done in the previous interaction. This is the situation in which every individual oscillates of his own accord, without generating fashions.
Fashion 1 0.5 0 2
0.9 0.94 4 6 8
0.90 10 12 14 m 16
0.86 18
200.82
Let us now move to the second effect: the impact of m on the length of fashion cycles. We define the length of a fashion cycle as the number of consecutive rounds in which one of the two strategies has been fashionable (i.e. has been adopted by more than 90 percent of the population). We have run simulations for several values of m ranging from 1 to 10, assuming a constant value of l ¼ 1U The pictures below represent the length of the fashion cycles (top) and their number (bottom). Increasing m clearly increases the length of fashion cycles and symmetrically reduces their number.
223
Fashion: Why People Like It and Theorists Do Not
Length
4 3 2 1 m 2
4
6
8
10
2
4
6
8
10
50
Number
40 30 20 10 m
5. COMMENTS: AN EXAMPLE Our model is quite stylized and cannot capture the full complexity of the variables in play; nonetheless it provides indications of the nature of fashion cycles. Both individual preferences and social interaction play a role in the model. Choices attach to one or another strategy according to their relative utility and novelty content; but, crucially, social interaction helps in uncovering what is new. More precisely, fashions depend on the size of social interaction l, and, given l, they disappear much more rapidly when novelty erodes quickly (m is low), more slowly when the appeal of novelty lasts longer (m is large). Real fashion cycles appear to conform to these properties. When we look at fashion trends, whether in furniture or dress, in cars or music, it is not difficult to see that styles or genres that are more complex and allow for different combinatory solutions change much more slowly than the more impermanent and fleeting features of fashions such as decoration, colors, and accessories.
224
LUCIANO ANDREOZZI AND MARINA BIANCHI
Take, for example, fashion trends in female dress. If we look at female dress styles, those involving the volumes, shapes, and proportions of a dress, we see that in the last century until the 1980s, styles changed with an almost regular cadence of 10 years, a regularity interrupted only by the two world wars. To have a visual image of these changes of silhouette, think of them in terms of selected letters of the alphabet. From 1900 until about 1908 the dominant silhouette was S-shaped, with dresses tightly fitted to the corseted body. In the next 10 years the outline became much more linear and the fit looser (an I-shape that sometimes, with a huge hat, became a T). In the twenties the I-shape was maintained and accentuated by flattening the bust and lowering the waistline to the hips, to form an H. The thirties saw dresses become tighter again, with padded shoulders dominating, so that the overall impression was that of a Y. This shape remained more or less unchanged till after the Second World War when Paris fashion creators re-launched a shape reminiscent of those from the end of the previous century, corsets squeezing the waist so as to create a body like a figure 8. This remained the dominant shape through the fifties. With the sixties, however, changes became more dramatic: geometrical shapes were dominant (A- O- and H-shapes) while the miniskirt transformed the whole look that became freer and informal. Tight I-shapes re-appeared in the seventies, padded shoulders in the eighties, but starting from these years changes in styles became less regular giving room to the co-presence of many different styles (see Tyrrell, 1986; Ewing, 2001; Glynn, 1978). Yet, if over the period up to the sixties styles showed regular longer trends, the same cannot be said for the other elements of dressing. Colors, patterns, accessories, and hairstyles changed more rapidly and often. Every season allowed for innumerable new variations that either helped identify a new style or prefigured a future change. If we look at the predominant fashion features in the famous roaring twenties, for example, we see that within the recognizably lean I-shape, changes in look were constant.14 At the beginning of the decade colors were mostly plain – beige, white, navy black – leaving brighter colors for the evening. Decorations, nonetheless, abounded, with scarves, beading, fringes, and embroidery being employed, and the hair made short and boyish. In the space of a few years, however, all these features had undergone change. The skirt, that reached its shortest length in 1927, had lengthened and was fitted more tightly to the hips. Colors became brighter, especially for summer. Patterns were now geometric, reflecting the influence of deco style in furniture. Undergarments molded rather than flattened. Cloche hats were larger and tilted back at the front to reveal rather than hide the face. The hair was worn longer with curls at the nape of the neck.15
Fashion: Why People Like It and Theorists Do Not
225
Finally, even the make-up went through changes: lipsticks became darker and eyes were blackened by kohl and mascara (see Tyrrell, 1986; Laver, 1995). What we learn from the example above is that trends in fashions have different lengths, longer waves mixing with shorter ones. Those element of fashion that are part of daily experience, that are more frequently seen and used and are easier (and cheaper) to grasp and experiment with, lower m and increase the frequency of changes in a way that does not happen in the case of styles that have longer life. This is much less true in more recent trends, however, especially for dresses. Nowadays, increased accessibility in terms of prices, information, and consumers’ skills has increased l and lowered m, with the result that fashion cycles tend to appear and disappear quickly not only at the periphery, in the minutiae and details of fashion, but also in what were once more stable elements of fashions, styles. Today it is stylistic pluralism that dominates, where the free and constant mixing and matching of styles has allowed for greater self expression and individuality (see Laver, 1995, p. 266).16
6. CONCLUSIONS Forms of consumption that express themselves in fashions are particularly challenging to theorists because they seem to defeat those explanations that are commonly used to analyze the emergence of habits, rules, and conventions. If we apply to fashion a purely rational addiction model, where people fall into consumption habits because of past consumption investments, we cannot explain switches in fashions – why some habits are abandoned in favor of new ones. Nor can we explain their timing, why some habits last longer and some others last less, or the fact that such shifts are synchronized among agents.17 On the other hand, if we adopt a pure coordination model, where convergence of interests helps in selecting an equilibrium solution, we cannot explain why such convergence suddenly becomes unstable and disappears. The models of fashion that have been produced in both the economic and sociological literature tend to solve the problem posed by the high degree of coordination shown by fashions by assuming that social preferences dominate individual preferences. People end up choosing the same goods because of their conformist preferences. Conversely, the task of explaining the ever-changing nature of fashion is simply shifted onto some prime movers: conspicuous consumers, upper classes, trendsetters, advertisers, and manufacturers.
226
LUCIANO ANDREOZZI AND MARINA BIANCHI
In our model we do without both these postulates. Individuals choose according to their ‘‘true’’ preferences whenever these are sufficiently strong that a clear preference over the alternatives is present. Secondly, we assume a homogeneous population in which there are no social classes, trendsetters, leaders, and followers. At each point in time, any agent can be a follower (when she imitates one of her fellows) or a leader (when she is imitated). Given these hypotheses in our approach to social interaction, even if in small degree, suffices to synchronize individuals’ oscillations into fashion cycles. The idea behind our model is that individual preferences and the desire for novelty drive fashion and thus make change an endogenous variable. What was an efficient choice when new, is no more so when novelty has vanished. Correspondingly, coordination and social interaction is integral to the fashion game in that these factors help the new to emerge through mutual emulation, experimentation, and learning.
NOTES 1. So permanent that we can even date different Roman statues by their hairstyles. 2. The first, and highly original, economic reflection devoted to the topic of fashion is the paper by Caroline Foley (1893). 3. See Section 2 for a discussion of this point. 4. These points are discussed in Bianchi (1998, 2003). 5. It is interesting to notice here that Georg Simmel (1957), whose name is frequently associated with Veblen’s when it comes to fashion, had a much more articulated approach to this topic. Like many who followed him, he argued that imitation in fashion follows a trickle down pattern, with only the upper classes as leaders of innovations. Yet he also discovered that fashion obeyed a desire for originality and self-expression, and opened social interaction to experimentation and change. A discussion of Simmel’s original contribution is in Sassatelli (2000). 6. For a detailed historical reconstruction of the debate on ‘‘conspicuous’’ consumption see Mason (1998). 7. Pesendorfer (1995) and Frijters (1998) modify these models, introducing the idea that luxury goods are provided by monopolists who face the classical problem of pricing a durable good (see also Gregor, 1948). They initially sell the good at a price higher than marginal cost to those who want to use it as a signal of wealth and status. Then they lower the price and sell it to the rest of the population. Once enough consumers have bought the good, and it no longer serves as a signal, they produce a new design of the same good that can be sold at a high price, and the cycle repeats itself. 8. See Glynn (1978, p. 59). 9. The spread of Internet technology among consumers has both accelerated and revealed the working of communities of consumers in discovering and sharing information. Von Hippel (2005) offers numerous examples, in sports, medical
Fashion: Why People Like It and Theorists Do Not
227
imaging equipment and information technology where innovations are driven by users rather than manufacturers. 10. On the role of Fechner’s studies on aesthetics see also Arnheim (1986, p. 40). 11. In a very anticipatory paper analysing experience-dependent choices, Richard Day (1986) introduces the concept of periodic variety and criticizes the traditional economic assumption of acyclic preferences. He shows how both auto-feedback and environment feedback processes have the effect of propagating irregularly changing states. 12. For a discussion of this point see Bianchi (2002). 13. An example might further clarify this point. Consider the difference between crime stories and classic works of literature such as J. Austen’s Pride and Prejudice or Shakespeare’s Hamlet. At a second reading (when you already known the culprit) a crime story has lost all its novelty appeal and is prone to bore, even though it was gripping on a first reading. On the contrary, classical novels such as Pride and Prejudice can be reread many times and still continue to be enjoyable. In our model, m would be very small (even one) for crime novels and quite large for Austen’s classic. In general, goods and activities that are complex and multi-dimensional tire less quickly than those whose characteristics are few and simple. 14. For women these were years of greater freedom and social mobility. Their active participation in work and sports, and the availability of new forms of entertainment such as motion pictures provided new social contacts and exposure. The fashions that emerged reflected these conquered freedoms of movement, activity, and aspirations. Indeed, many of their features were directly imported from the world of sports and work (tweeds and pullovers, shorts, and pants). Meanwhile, technological improvements and innovation in the production of fabric made fashion more accessible and democratic. Clothes required, in fact, little material and had an easy cut, stockings had become cheaper, decoration was easily altered and modular, thus allowing for constant alternations and changes. 15. Haircuts remained short in these years but with many changes. The ‘‘bob’’ of the early twenties was first abandoned in favor of the ‘‘shingle,’’ a cut with naturallooking waves, and later in favor of a short and masculine cut – the Eton crop. 16. In dresses, this is the period of the triumph of ethnic, hippie, punk, grunge, and mix-match looks, of revivals of the thirties, forties, and fifties styles, with actual vintage clothes being used. 17. For example, Karni and Schmeidler dismiss this kind of explanation on fashion cycles on the ground that ‘‘while changing tastes due to habit formation may explain certain aspects of variations in individual demands over time, it does not explain the correlated changes in individual demand that constitute a fad’’ (Karni & Schmeidler, 1990, p. 262).
REFERENCES Apter, M. J. (Ed.) (2001). Motivational styles in everyday life. A guide to reversal theory. Washington, DC: American Psychological Association. Arnheim, R. (1986). New essays on the psychology of art. Berkeley: University of California Press.
228
LUCIANO ANDREOZZI AND MARINA BIANCHI
Bagwell, L. S., & Bernheim, B. D. (1996). Veblen effects in a theory of conspicuous consumption. American Economic Review, 86, 349–373. Barkow, J. H., Cosmides, L., & Tooby, J. (Eds). (1992). The adapted mind. Evolutionary psychology and the generation of culture. Oxford: Oxford University Press. Becker, G. S. (1996). Accounting for tastes. Cambridge, MA: Harvard University Press. Becker, G. S., & Murphy, K. (1988). A theory of rational addiction. Journal of Political Economy, 96, 675–700. Berlyne, D. E. (1971). Aesthetics and psychobiology. New York: Appleton Century Crofts. Berlyne, D.E., & Madsen, K.B. (Eds). (1973). Pleasure, reward, preference. New York: Academic Press. Bianchi, M. (1998). Taste for novelty and novel tastes. The role of human agency in consumption. In: M. Bianchi (Ed.), The active consumer. Novelty and surprise in consumer choice (pp. 64–86). London: Routledge. Bianchi, M. (2002). Novelty, preferences, and fashion: When goods are unsettling. Journal of Economic Behavior and Organization, 4, 1–18. Bianchi, M. (2003). A questioning economist: Tibor Scitovsky’s attempt to bring joy into economics. Journal of Economic Psychology, 24, 391–407. Blume, L. E. (1995). The statistical mechanics of best-response strategy revision. Games and Economic Behavior, 11(2), 111–145. Camerer, C., Lowenstein, G., & Prelec, D. (2005). Neuroeconomics: How neuroscience can inform economics. Journal of Economic Literature, 44, 9–64. Congleton, R. D. (1989). Efficient status seeking externalities and the evolution of status games. Journal of Economic Behavior and Organization, 11, 175–190. Conlisk, J. (2003). Dynamic preferences and specialized tastes. Economics Letters, 80, 357–364. Corneo, G., & Jeanne, O. (1999). Segmented communication and fashionable behavior. Journal of Economic Behavior and Organization, 39, 371–385. Cowan, R., Cowan, W., & Swann, G. M. P. (2004). Waves in consumption with interdependence among consumers. Canadian Journal of Economics, 37(1), 149–177. Crozier, W. R., & Chapman, A. J. (Eds). (1984). Cognitive processes in the perception of art. Amsterdam: North Holland. Day, R. H. (1986). A note on endogenous preferences and adaptive economizing. In: R. H. Day & G. Eliasson (Eds), The dynamics of market economies (pp. 153–170). Amsterdam: North-Holland. De Vany, A. (2004). Hollywood economics. How extreme uncertainty shapes the film industry. London: Routledge. Elster, J. (1989). Social norms and economic theory. Journal of Economic Perspectives, 3(4), 99–117. Ewing, E. (2001). History of twentieth century fashion. London: Batsford. Foley, C. A. (1893). Fashion. The Economic Journal, 3, 458–474. Frank, R. H. (1985). Choosing the right pond. Oxford: Oxford University Press. Frijters, P. (1998). A model of fashion and status. Economic Modelling, 15, 501–517. Glynn, P. (1978). In fashion: Dress in the twentieth century (illustrated by Madeleine). New York: Oxford University Press. Gregory, P. (1948). Fashion and monopolistic competition. Journal of Political Economy, 56, 69–75. Herrnstein, R., & Prelec, D. (1992). Melioration. In: G. Loewenstein & J. Elster (Eds), Choice over time (pp. 235–264). New York: Russell Sage Foundation.
Fashion: Why People Like It and Theorists Do Not
229
Jones, S. R. G. (1984). The economics of conformism. Oxford: Blackwell. Kaplan, S. (1992). Environmental preferences in a knowledge-seeking, knowledge-using Organism. In: J. H. Barkow, L. Cosmides & J. Tooby (Eds), The adapted mind. Evolutionary psychology and the generation of culture (pp. 581–598).Oxford: Oxford University Press. Karni, E., & Schmeidler, D. (1990). Fixed preferences and changing tastes. American Economic Review, 80, 262–267. Laver, J. (1995, rev., exp., and updated [1969]). Costume and fashion. London: Thames and Hudson. Leibenstein, H. (1950). Bandwagon, snob, and veblen effects in the theory of consumers’ demand. Quarterly Journal of Economics, 64, 183–207. Mason, R. (1998). The economics of conspicuous consumption. Theory and thought since 1700. Cheltenam: Edward Elgar. Pesendorfer, W. (1995). Design innovation and fashion cycles. American Economic Review, 85, 771–792. Pollak, R. A., & Wales, T. J. (1992). Demand system specification and estimation. Oxford: Oxford University Press. Robinson, D. E. (1961). The economics of fashion demand. The Quarterly Journal of Econo mics, 75, 376–398. Sassatelli, R. (2000). From value to consumption. A social-theoretical perspective on Simmel’s Philosophie des Geldes. Acta Sociologica, 4(3), 207–218. Scitovsky, T. (1992 [1976]). The joyless economy: The psychology of human satisfaction (revised edition). Oxford: Oxford University Press. Simmel, G. (1957). Fashion. American Journal of Sociology, 62(6), 541–558. Tilman, R. (1998). Georg Simmel and Thorstein Veblen on fashion Fin de Sie´cle. In: W. Samuels (Ed.), The founding of institutional economics, the leisure class, and sovereignty (pp. 282–301). London: Routledge. Tyrrell, A. V. (1986). Changing trends in fashion: Patterns of the twentieth century, 1900–1970. London: Batsford. Von Hippel, E. (2005). Democratizing innovation. Cambridge, MA: MIT Press. Young, P. (1998). Individual strategy and social structure: An evolutionary theory of institutions. Princeton, NJ: Princeton University Press.
This page intentionally left blank
DOES CONTEXT MATTER MORE FOR SOME GOODS THAN OTHERS? Robert H. Frank ABSTRACT Context is known to affect evaluation for many goods. For example, a house of any given size is more likely to be viewed as adequate the larger it is relative to other houses in the same locale. If evaluations of some goods are more sensitive to context than others, there is no presumption that privately optimal consumption patterns will be socially optimal. Rather, consumers will spend too much on goods whose evaluations depend most strongly on context and too little on those whose evaluations depend least strongly on context. For instance, if evaluations of houses are more sensitive to context than evaluations of leisure, then people will spend too much money on houses and too little time with family and friends. But if context sensitivity is the same for all goods, no distortions result. This paper suggests theoretical grounds for expecting context sensitivity to differ across goods. Evaluations should be more sensitive to context for goods whose consumption is more readily observed by others and also for goods for which relative consumption is linked to other important payoffs. The quality of school that a child attends, for example, is often strongly linked to its parents’ relative expenditures on housing. A survey of empirical evidence suggests that observed differences in context sensitivity track the differences predicted on theoretical grounds.
The Evolution of Consumption: Theories and Practices Advances in Austrian Economics, Volume 10, 231–248 Copyright r 2007 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 1529-2134/doi:10.1016/S1529-2134(07)10010-7
231
232
ROBERT H. FRANK
In traditional economic models, individual utility depends only on absolute consumption. Recent years, however, have seen renewed interest in economic models in which individual utility depends also on relative consumption. In contrast to traditional models, these models identify a fundamental conflict between individual and social welfare. The conflict stems from the fact that concerns about relative consumption are stronger in some domains than in others. The disparity gives rise to expenditure arms races focused on positional goods – those for which relative position matters most. The result is to divert resources from nonpositional goods, causing welfare losses. In this paper, I begin by noting that the very definition of a positional good is contextual. Next, I briefly describe the conditions that give rise to expenditure arms races. I then examine general theoretical and empirical evidence relevant to the claim that relative position is important. Finally, I discuss what theoretical considerations and empirical evidence have to say about the extent to which positional concerns affect the individual valuations of specific goods.
1. DEFINING POSITIONAL AND NONPOSITIONAL GOODS To help fix ideas, I begin with two simple thought experiments. In each, you must choose between two worlds that are identical in every respect except one. The first choice is between world A, in which you will live in a 4000square-foot house and others in 6000-square-foot houses; and world B, in which you will live in a 3000-square-foot house, others in 2000-square-foot houses. Once you choose, your position on the local housing scale will persist. If only absolute consumption mattered, A would be clearly better. Yet most people say they would pick B, where their absolute house size is smaller but their relative house size is larger. Even those who say they would pick A seem to recognize why someone might be more satisfied with a 3000square-foot house in B than with a substantially larger house in A. In the second thought experiment, your choice is between world C, in which you would have four weeks a year of vacation time and others would have six weeks; and world D, in which you would have two weeks of vacation, others one week. This time most people pick C, choosing greater absolute vacation time at the expense of lower relative vacation time. I use the term positional good to denote goods for which the link between context and evaluation is strongest and the term nonpositional good to
Does Context Matter More for Some Goods than Others?
233
denote those for which this link is weakest.1 In terms of the two thought experiments, housing is thus a positional good, vacation time a nonpositional good. The point is not that absolute house size and relative vacation time are of no concern. Rather, it is that positional concerns weigh more heavily in the first domain than in the second.
2. THE CONFLICT BETWEEN INDIVIDUAL AND COLLECTIVE INTEREST When the strength of positional concerns differs across domains, the resulting conflict between individual and social welfare is structurally identical to the one inherent in a military arms race. To illustrate, consider rival nations faced with deciding how to apportion available resources between domestic consumption and military armaments. Each country’s valuations are typically more contextdependent in the armaments domain than in the domain of domestic consumption. After all, having lower domestic consumption than one’s rival might entail psychological discomfort, but being less well-armed could spell the end of political independence. The familiar result is a mutual escalation of expenditure on armaments that does not enhance security for either nation. Because the extra spending comes at the expense of domestic consumption, its overall effect is to reduce welfare. Note that if each country’s valuations were equally context-sensitive in the two domains, there would be no arms race, for in that case the attraction of having more arms than one’s rival would be exactly offset by the penalties of having lower relative consumption. For parallel reasons, the modal responses to the two thought experiments just discussed suggest an equilibrium in which people consume too much housing and too little leisure.2 In contrast, conventional welfare theorems, which assume that individual valuations depend only on absolute consumption, imply optimal allocations of housing and leisure. Is this default assumption a reasonable one? I consider this question from both theoretical and empirical perspectives.
3. THE NATURE OF THE UTILITY FUNCTION: THEORETICAL CONSIDERATIONS No serious scientist denies that animal nervous systems were forged by natural selection. In the Darwinian view, animal drives were selected for
234
ROBERT H. FRANK
their capacity to motivate behaviors that contribute to reproductive success. Reproductive success, in turn, is fundamentally about resource acquisition: other things equal, the more resources an animal has, the more progeny it leaves behind. What matters is not the absolute number of offspring an individual has, but rather how its progeny compare in number with those of other individuals. A specific trait will thus be favored by natural selection less because it facilitates resource acquisition in absolute terms than because it confers an advantage in relative terms. Frequent famines were an important challenge in early human societies. But even in the most severe famines, there was always some food. Those with relatively high resource holdings got fed, while others often starved. On the plausible assumption that individuals with the strongest concerns about relative resource holdings were most inclined to expend the effort necessary to achieve high rank, such individuals would have been more likely than others to survive food shortages. Relative resource holdings were also important in implicit markets for marriage partners. In most early human societies, high-ranking males took multiple wives, leaving many low-ranking males with none. Even in contemporary societies, sexual attractiveness is strongly linked to relative resource holdings. So here, too, theory predicts that natural selection will favor individuals with the strongest concerns about relative resource holdings. The motivational structure expected on the basis of theoretical considerations is thus consistent with the modal choice patterns in our two thought experiments. Evolutionary theory also helps identify the specific reference groups that are likely to matter most. In evolutionary terms, falling behind one’s local rivals can be lethal, whereas comparisons with others who are distant in time or space are typically irrelevant. And as the empirical studies mentioned below confirm, it is local rank that matters most.
4. THE NATURE OF THE UTILITY FUNCTION: EMPIRICAL EVIDENCE The hypothesis that concerns about local rank are part of the evolved circuitry of the human brain is supported by evidence of specific neurophysiological processes that respond to local relative position. For example, local rank appears to affect, and be affected by, concentrations of the neurotransmitter serotonin, which regulates moods and behavior. Within limits,
Does Context Matter More for Some Goods than Others?
235
elevated serotonin concentrations are associated with enhanced feelings of well-being. (The drug Prozac, widely prescribed for depression and other mood disorders, increases the effective concentrations of serotonin in the brain.) In males, concentrations of the sex hormone testosterone appear to have a similar relationship with local rank. Reductions in local rank tend to be followed by reductions in plasma testosterone levels, whereas these levels tend to rise following increases in rank. A player who wins a tennis match decisively, for example, experiences a post-match elevation in plasma testosterone, and his vanquished opponent experiences a post-match reduction. As with serotonin, there is some evidence that elevated concentrations of testosterone facilitate behaviors that help achieve or maintain high local rank.3 Further evidence of the importance of relative position comes from studies of the determinants of happiness, or subjective well-being. Investigators find that whereas average happiness levels within a country tend to be highly stable over time, even in the face of significant economic growth, individual happiness levels within any country at a given moment of time depend strongly on income (Easterlin, 1995). Recent work employing richly detailed panel data further confirms the importance of local comparisons. This work documents a robust negative association between individual happiness measures and average neighborhood income, a link that does not appear to stem from selection effects (Luttmer, 2005). The hypothesis that local rank matters also has testable implications for the distribution of wages within firms (Frank, 1984). If some value high local rank more than others, then economic surplus is maximized by having workers sort themselves into separate firms in accordance with their respective valuations. Within each firm, the equilibrium distribution of wages will be more compressed than the corresponding distribution of marginal products. In effect, the labor market serves up compensating wage differentials for local rank, much as it does for other nonpecuniary employment conditions. This pattern, which is widely observed, is inconsistent with models in which local rank has no value.
5. HOW POSITIONAL CONCERNS VARY ACROSS CATEGORIES: THEORETICAL CONSIDERATIONS The Darwinian perspective on human motivation suggests that positional concerns might be expected to vary in accordance with the extent to which
236
ROBERT H. FRANK
relative consumption in different categories contributes to inclusive fitness. Some examples follow.
5.1. Leisure Consider the tradeoff faced by each individual between increased consumption of leisure, on the one hand, and increased acquisition of material resources on the other. When threats to survival are acute, as during famines, those who stand high in the distribution of material resources will invariably get enough to eat. In contrast, those who emphasize leisure over material resource acquisition often starve. So even though everyone might enjoy greater health and well-being if all consumed more leisure, it may not be advantageous for individuals to consume more leisure unilaterally. Again, the problem is that limited food stocks in famines are more likely to go to those with high relative incomes than to those with high relative leisure.
5.2. Environmental Amenities The same considerations that suggest that leisure should rank low on the positionality scale suggest a similarly low ranking for other nonmaterial consumption amenities, such as freedom from noise and pollution. By the same token, workplace amenities such as grievance procedures, additional variety and comfort features should weigh less heavily in positional competition than the wage income that must be sacrificed to obtain them.
5.3. Investment in Children To have raised offspring that are well equipped to compete in their cohort is one of the most conspicuous yardsticks by which success is measured in the Darwinian framework. This task is fruitfully viewed as a contest. Most parents, for example, want their children to hold interesting, well-paying jobs some day, but such adjectives are inherently context-dependent. Thus an interesting job is simply one that is more interesting than most other jobs. As in virtually every contest, contestants attempting to launch their children well in life take a variety of steps to outdo or keep pace with their rivals. Accordingly, categories of expenditure that contribute to this goal, such as expenditures on schooling, should be highly positional.
Does Context Matter More for Some Goods than Others?
237
5.4. Visibility By their very nature, concerns about position cannot focus on a given category unless relative consumption in that category can be measured. Other things equal, categories of consumption that are not readily observed should thus be relatively less positional. But observability is a necessary, not sufficient, condition for positionality. The fact that someone is a smoker can be readily observed by others, for example, but that does not make smoking a positional good.
5.5. Safety and Insurance Many expenditures, such as those on accident prevention and insurance, yield benefits only in states of the world that occur with very low probability. As Arthur Robson and others have argued, the Darwinian perspective on human motivation suggests a risk-seeking posture toward such expenditures, particularly for males (Robson, 1992). As noted earlier, most human societies have been polygynous, and in such societies the highestranking males typically sire a disproportionate share of all offspring. This skewed payoff structure encourages high-variance strategies. The analogy is often cited to a football team that finds itself substantially behind in the fourth quarter. If it sticks to a low-variance running strategy, it is almost sure to lose. But if it switches to a high-variance passing game, it creates at least a slim chance of winning. The fact that expenditures on safety and insurance are also relatively difficult for others to observe reinforces the prediction that these categories will be nonpositional.
5.6. Signals of Ability Within specific labor markets, income and ability tend to be strongly correlated. And because observable consumption and income are also strongly correlated, observable consumption will often be a crude but effective predictor of ability. To the extent that it is individually advantageous to be seen by others as someone with high ability, the tendency will be to steer expenditures in favor of consumption categories that signal high ability. The prediction is that expenditures in these categories will run higher in environments in which independent measures of ability are less readily available.
238
ROBERT H. FRANK
5.7. Savings Savings might also be predicted to be nonpositional on grounds of unobservability. Yet if reduced savings today means reduced capacity to spend on positional consumption in the future, the mere fact that current savings is unobservable is not decisive. Recent work in behavioral economics has identified a general tendency to discount future costs and benefits much more heavily than assumed in traditional economic models. Perhaps the problem is that whereas the current consequences of savings decisions can be experienced directly, their future consequences must be imagined. If so, then we would expect savings to be nonpositional. Alternatively, it may be that expenditures early in life are inherently more positional than those occurring later. Suppose, for example, that a parent must choose between putting money aside to support a comfortable standard of living during retirement or using that same money toward a down payment on a house in a better school district. As noted earlier, expenditures on school quality are predicted to be highly positional. And since these expenditures occur early in life, the prediction is that their positional nature will tend to crowd out savings. 5.8. Public Goods The definition of a public good entails two features: It is nonexcludable, meaning that it is difficult or impossible to exclude people from consuming it once it is produced; and it is nonrival, meaning that one person’s consumption of it does not diminish the amount of it available for consumption by others. To the extent that goods such as public parks are equally available to all consumers, they are inherently nonpositional.
6. HOW POSITIONAL CONCERNS VARY ACROSS CATEGORIES: EMPIRICAL EVIDENCE In this section, I consider a variety of empirical evidence that bears on the validity of the predictions just discussed. Some of this evidence comes in the form of surveys that ask respondents to respond to questions like the ones posed in the two thought experiments. Some comes in the form of econometric studies of expenditure patterns. Some comes in the form of observations about how changes in income inequality appear to affect
Does Context Matter More for Some Goods than Others?
239
expenditures patterns. And finally some consists of observations about the specific ways that society has chosen to regulate different expenditure categories. 6.1. Leisure Sara Solnick and David Hemenway have conducted several surveys in which they ask participants to choose between hypothetical worlds in the manner illustrated in the two thought experiments discussed at the outset (Solnick & Hemenway, 1998, 2005). Response patterns in these surveys consistently reveal leisure to be near the bottom of the positionality scale. In the same spirit, Renee Landers, James Rebitzer, and Lowell Taylor asked associates in large law firms which they would prefer, their current situation, or an otherwise similar one with an across-the-board cut of 10 percent in both hours and pay (Landers, Rebitzer, & Taylor, 1996). By an overwhelming margin, respondents chose the latter. Similar results have been found in other countries. In a recent survey, for example, Swedish respondents considered income more positional than leisure (Carlsson, Johansson-Stenman, & Martinsson, 2003). Concerns about position have also been shown to affect labor force participation – in some studies by much more than such traditional factors as local wage and unemployment rates. David Neumark and Andrew Postlewaite found, for example, that a woman whose sister’s husband earned more than her own husband was 16–25 percent more likely than others to seek paid employment (Neumark & Postlewaite, 1998). Changes in the distribution of income provide yet another opportunity to test for the presence of positional concerns. One of the core findings of behavioral economics is loss aversion, the tendency for the pain caused by a loss of given magnitude to be greater than the pleasure caused by a gain of the same size. When income inequality increases, the expectation is thus that the pain experienced by those who fall behind is greater than the pleasure experienced by those who pull ahead. If leisure is less positional on average than other categories of consumption, it then follows that a rise in income inequality will cause a net increase in hours worked, as those who have fallen behind attempt to undo the injury they have experienced. Samuel Bowles and Yongjin Park found in a recent study that total hours worked, both across countries and over time within countries, are in fact positively associated with higher earnings inequality (Bowles & Park, 2005). Models that incorporate positional concerns predict these links.4 Traditional labor market models do not.
240
ROBERT H. FRANK
As discussed earlier, if positional concerns differ across categories, expenditure arms races focused on positional goods will produce a welfarereducing diversion of resources from nonpositional goods. So, if leisure is less positional than other categories of consumption on average, the tendency will be for people to consume too little leisure and spend too much on other categories of consumption. The categories of behavior society chooses to regulate reflect social judgments about the relevant deficits and excesses. In the case of leisure, almost all countries encourage leisure consumption through regulation and social norms. Long before governments became involved, religions attempted to encourage leisure consumption by designating Sabbath days on which work was forbidden. In the United States, the Fair Labor Standards Act encourages shorter working hours by its provision requiring premium pay for labor performed in excess of 8 hours per day, 40 hours per week, or on national holidays. European regulations are even stricter in their support of shorter hours. And many jurisdictions continue to enforce Blue laws, which make it unlawful for establishments to remain open during certain periods. Regulations are data. If so many countries actively intervene to constrain the number of hours that people would otherwise choose to work in unregulated markets, it must be because people believe that working longer hours would reduce welfare. The conventional explanations offered for these regulations have been shown inadequate. Thus, although France defended its recent requirement of a 35-hour workweek on the grounds that it was needed to stimulate jobs, officials could cite no credible evidence for the existence of such stimulus. Similarly, although many have defended hours regulations as needed to protect workers from employers with market power, the constraints imposed by such regulations bind much more heavily for hourly workers in low-wage labor markets, which are among the most highly competitive by conventional yardsticks. Salaried workers in highwage labor markets are relatively unconstrained by hours regulations, even though their employers are much more likely to occupy dominant market positions. The observed patterns of regulation are consistent with the hypothesis that leisure gets short shrift because of positional concerns. 6.2. Environmental Amenities Labor legislation in countries around the globe regulates not just hours but also a variety of other aspects of the labor contract. In many places, the law mandates specific workplace grievance procedures, and some countries have adopted statutes that attempt to make the workplace more democratic.
Does Context Matter More for Some Goods than Others?
241
These regulations are also consistent with the hypothesis that such amenities would otherwise be underprovided because of positional concerns.
6.3. Investment in Children If the goal of launching one’s children well in life is fruitfully viewed as a contest, then one particularly sensitive step in that contest is the decision about when a child should begin school. Looking ahead, parents know that applicants are admitted to the most selective universities on the basis of having performed well relative to their classmates. Having one’s child start kindergarten a year later than others would thus confer an advantage, because it would make the child bigger, stronger, and more intelligent relative to his classmates. But because other parents could easily respond by holding their own children back, the unregulated equilibrium might well be one in which most children did not start kindergarten until 8 or 9 years of age. Socially, that outcome is clearly inefficient. And most jurisdictions have thus enacted laws mandating school attendance for 6-year olds. Viewed as data, these regulations are consistent with the hypothesis that investments in children are highly positional. In the United States, one of the most important investments a family can make in its children’s future is to buy a house in a good school district. As a general rule, the quality of a neighborhood school is strongly correlated with the average price of houses in the neighborhood. This is true in part because local property taxes are a major source of school funding. But because of the importance of peer effects in the classroom, the better schools tend to be located in more expensive neighborhoods even in countries in which school budgets are independent of local property taxes. Yet no matter how much every family spends on housing, the inescapable mathematical logic of musical chairs ensures that half of all children will attend schools in the bottom half of the school quality distribution. There is considerable evidence for the existence of bidding wars for houses in the best school districts. As Elizabeth Warren and Amelia Warren Tyagi have shown, for example, most of the extra income earned by families as a result of the move to two-earner couples was consumed by higher housing prices as these families sought to buy homes in safer neighborhoods with better schools (Warren & Tyagi, 2003). Warren and Tyagi also present evidence that each time the credit industry relaxed its terms by permitting lower down payments and longer payoff periods for home mortgages, the primary effect was again a bidding war for these same preferred neighborhoods.
242
ROBERT H. FRANK
Additional evidence for such bidding wars comes from how housing expenditures have responded to changes in the distribution of income. The permanent income and life-cycle theories of consumption, which assume that each family’s expenditure on housing is independent of what others spend, predict no change in the median household’s spending on housing in response to large income gains at the top of the income distribution. Recent large changes in the distribution of income allow us to test that prediction. Although the top 1 percent of earners in the United States today earn about three times as much as in 1979, the median household now earns only about 15 percent more.5 If we assume, conservatively, that expenditure on housing is proportional to income, then the median earner should be spending roughly 15 percent more on housing than in 1979. In contrast, models that incorporate positional concerns predict that sharply increased spending by top earners will exert indirect upward pressure on spending by the median earner. When top earners build larger houses, for example, they shift the frame of reference that defines what others slightly below them on the income scale consider an acceptable or desirable house. And when those people respond by building bigger houses, they in turn shift the frame of reference for those just below them, and so on, all the way down. Thus the median size of a newly constructed house, which stood at less than 1600 square feet in 1980, had risen to over 2100 square feet by 2001 – more than twice the increase predicted by the permanent income and life cycle theories.6 Indirect evidence supports the view that increased housing expenditures by the median household are at least in part a consequence of increased income inequality. For example, U.S. counties that experienced larger increases in earnings inequality also experienced significantly larger increases in personal bankruptcy rates, divorce rates, and average commute times (Frank & Levine, 2005). If regulations can be viewed as data that shed light on the nature of positional concerns, the same may be true of the ways in which we choose to finance educational services. The hypothesis that investment in education of children is highly positional predicts that we should find it attractive to adopt mechanisms for paying for education that discourage expenditure arms races. The principal education finance schemes employed in the United States include just such a design feature. Most jurisdictions levy a tax that entitles children to ‘‘free’’ public education. Parents also have the option of purchasing private education, but those who do so must continue to pay their school tax. To improve upon the option of sending one’s child to the public schools, a family must essentially forfeit its entitlement to free educational services and start purchasing
Does Context Matter More for Some Goods than Others?
243
educational services from scratch in the private sector. This arrangement creates a sharp disincentive to spending more than the per-pupil allotment specified in the public school budget.7 We could have chosen different arrangements. Under a voucher system of educational finance, for example, a family could boost the amount of educational services it purchased for its children without forfeiting the amount it had paid in school taxes. Under such a system, many families would undoubtedly find it attractive to give their children a little more than called for in the basic public plan. But again, the concept of a ‘‘good education’’ is context-dependent. As families responded to the incentive to spend more, a side effect would be to devalue the education received by others, thus imposing pressure on them to increase their spending as well. Except for the possible fact that it might encourage an expenditure arms race, a voucher scheme appears attractive in numerous other respects. For example, voucher proponents have argued that the system would stimulate quality improvements and cost reductions of the sort generally associated with increased competition in other sectors. Distributional objections to vouchers could easily be addressed by making the vouchers progressive. That we have nonetheless rejected the voucher approach is consistent with the hypothesis that our current method of educational finance, for all its flaws, has the important virtue of keeping educational expenditures under control. 6.4. Visibility In a 2005 paper, Ori Heffetz has attempted to test the hypothesis that the observability of an expenditure category predicts the extent to which valuations in that category are positional (Heffetz, 2005). On the basis of a detailed telephone survey, Heffetz assigned a visibility index, or ‘‘vindex,’’ to more than 30 categories of expenditure recorded by the Consumer Expenditure Survey. Categories with the highest vindex values included cars, jewelry, and clothing, while those with the lowest visibility included car insurance, life insurance, and household utilities. Heffetz found that the more visible a good is, the more likely it is to be a luxury item. The share of a household’s income spent on cars and jewelry, for instance, goes up as income rises, while budget shares devoted to insurance and utilities go down as income rises. 6.5. Safety and Insurance In surveys involving students at Cornell University, I have substituted the following thought experiment for the second thought experiment discussed
244
ROBERT H. FRANK
at the outset of this paper: ‘‘Your choice is between world C, in which you would have a 2 in 100,000 chance of being killed on the job each year, and others would have a 1 in 100,000 chance of being killed; and world D, in which you would have a 4 in 100,000 chance of being killed on the job each year, while others would have an 8 in 100,000 chance of being killed. Here again, the overwhelming majority pick C, indicating a preference of greater absolute safety time at the expense of lower relative safety. Similar results have been found elsewhere. Thus, in the Swedish survey cited earlier, respondents considered the monetary value of a company car more positional than its safety (Carlsson et al., 2003). The hypothesis that expenditures on safety and insurance rank low on the positionality scale implies a tendency for unregulated individual expenditures in these categories to fall short of their respective socially optimal values. And most societies have enacted various forms of legislation whose effect to increase expenditures in these categories. In the case of workplace safety regulation, the claim is again heard that such measures are needed to protect workers from being exploited by employers with significant market power. But here, too, the tendency is for safety regulations to have their greatest impact in labor markets that by most measures are among the most highly competitive. This pattern is consistent with the observation that these regulations serve as a positional arms control agreement – one that insulates workers not from exploitation by employers with market power, but rather from the consequences of unbridled competition for position.8
6.6. Signals of Ability In some occupations, the correlation between income and the underlying abilities that are valued most highly is much larger than in others. Among trial lawyers, for example, the link between income and ability is much stronger than it is in the case of university professors within any given discipline. Expenditures on such items as cars, clothing, and jewelry should thus serve as more effective signals of ability for lawyers than for professors. This observation suggests that lawyers with a given income will spend more on cars and clothing than will professors with that same income. A related prediction is that the difference between the two groups will be greater in large cities than in small cities, since other sources of information about ability is likely to be more easily accessible in small cities. I know of no specific study that has attempted to test these predictions.
Does Context Matter More for Some Goods than Others?
245
7. SAVINGS Any successful consumption theory must accommodate three stylized facts: (1) as income grows over time, savings rates remain roughly constant; (2) consumption is more stable over time than income; and (3) highincome persons save at greater rates than low-income persons. James Duesenberry’s relative income hypothesis, which holds that a family’s savings rate is an increasing function of both its position in its local comparison group and its own previous peak consumption, has consistently tracked these facts (Duesenberry, 1949). In contrast, the competing life cycle and permanent income hypotheses accommodate them only through tortured ad hoc modifications. For example, the higher savings rates of persons with higher permanent incomes are ‘‘explained’’ by positing a bequest motive for rich consumers. But why shouldn’t the American poor, who have much higher incomes than most of the world’s citizens, also want to leave bequests? They undoubtedly do, Duesenberry would answer, but because of their relatively disadvantageous position, they have other more pressing needs to attend to. Another problem for the permanent income hypothesis is that, contrary to Milton Friedman’s assertion that windfall income will be mostly saved, people actually consume windfall income at almost the same rate as permanent income. To this observation, Friedman responded that consumers appear to have unexpectedly short planning horizons. But if so, then consumption doesn’t really depend primarily on permanent income. In addition to tracking the three main stylized facts of consumption data, Duesenberry’s model makes numerous other detailed predictions. In the highly segregated neighborhoods that prevailed when Duesenberry wrote, a black family with a given income would have higher income relative to its neighbors than a white family with the same income. And Duesenberry’s theory correctly predicted that black families will save at higher rates than white families with the same income. The permanent income hypothesis and the life cycle hypothesis, both of which disavow any role for context in consumption decisions, predict that families will save at the same rate irrespective of race.
7.1. Public goods According the median voter theorem, a public project will be enacted if it meets the approval of the median voter. As noted earlier, the median household earns about 15 percent more today in real terms than in 1979. So to the
246
ROBERT H. FRANK
extent that willingness to pay for public goods is an increasing function of absolute income, as in traditional economic models, basic public services should be provided at a higher level today than in 1979. Models that incorporate positional concerns make different predictions. As noted, these models stress that the sharp increases in income at the top of the income ladder have imposed growing financial distress on earners in the middle. Positional models thus predict a reduced willingness by the median voter to fund additional public services or even to pay for existing ones. The median voter might know that we would all do better if our children attended better funded schools; that we would all do better if we repaired the potholes in the roads in a timely fashion; and that we would all do better if our municipal water supplies were free of disease-causing microbes and toxic heavy metals. Yet public budgets in the United States have been steadily shrinking in all these areas.9 Public school teachers, who earned 119 percent of the average college graduate’s salary in 1962, now earn less than the average college graduate. The Scholastic Achievement Test scores and class rank of persons entering public school teaching have declined steadily during the same period. Class sizes have been growing larger steadily. Some 40 percent of American roads are in backlog, meaning that they are overdue for maintenance. Our bridges are in a similar state of disrepair. The deaths of 10 motorists when a bridge collapsed over Schoharie Creek on Interstate 90 in New York prompted an emergency inspection of all the state’s bridges. More than a third were found to be structurally compromised by deferred maintenance. Roads in backlog cost from two to five times as much to repair as those maintained on schedule. And in the meantime, damaged road surfaces cause an average of $120 of damage each year to every car and truck in the country. Cutbacks have also occurred in public health. Government inspections of meat processing plants, for example, now occur at only 25 percent of the rate they did in the 1980s, despite the emergence of E. coli and other dangerous new threats. We have been closing drug treatment and prevention centers, despite evidence that we prevent $7 in damage for every dollar we spend on these programs. More than 40 million Americans are now served by municipal water systems from whose antiquated pipes lead and manganese leach into their drinking water. These public spending cutbacks are inconsistent with traditional economic models, which assume that valuations depend only on absolute income. But if relative income also matters, as all evidence suggests, then the observed cutbacks are a predicted consequence of increased dispersion in the distribution of income.
Does Context Matter More for Some Goods than Others?
247
8. CONCLUDING REMARKS The idea that context affects evaluation is neither new nor controversial. As a young man just out of college, I served for two years as a Peace Corps Volunteer teacher in a small village in Nepal. The house I lived in there had a grass roof that leaked when it rained hard. It was an extremely small house by U.S. standards; it had no plumbing and no electricity. If you lived in such a house here, your children would be embarrassed to bring their friends home. Yet never once did I feel embarrassed about living in that house in Nepal, because it was actually a perfectly satisfactory house in that context. There is persuasive theoretical and empirical evidence for believing that evaluations of some goods are more sensitive to context than others. As discussed, for example, evaluations of houses are almost surely more sensitive to context than evaluations of leisure. In light of such differences, there is simply no presumption that Adam Smith’s invisible hand will deliver an optimal mix of goods and services. On the contrary, we will tend to spend too much on goods whose evaluations depend most strongly on context and too little on those whose evaluations depend least strongly on context.
NOTES 1. The late Fred Hirsch (1976), coined these terms. 2. For a formal demonstration of this result, see Frank (1985). 3. For a review of studies of the relationships between local rank, serotonin and testosterone, see Frank, (1999, Chapter 9). 4. These models also predict the observed negative relationship between income inequality and average happiness levels (see Alesina, McCulloch, & Tella, 2001). 5. www.inequality.org. 6. http://www.census.gov/prod/2003pubs/02statab/construct.pdf; http://www.census. gov/hhes/income/histinc/f03.html. 7. For a more formal description of how this kind of public finance scheme affects spending incentives, see Frank (1995). 8. For more on this point, see Frank, (1991). 9. For a more detailed discussion of the claims in the following paragraphs, see Frank, (1999, Chapter 4).
REFERENCES Alesina, A., McCulloch, R., & Tella, R. D. (2001). Inequality and happiness: Are Europeans and Americans different? CEPR Discussion Paper No. 2877, July.
248
ROBERT H. FRANK
Bowles, S., & Park, Y. (2005). Emulation, inequality, and work hours: Was Thorstein Veblen right? Economic Journal, 115(507), F397–F412. Carlsson, F., Johansson-Stenman, O., & Martinsson, P. (2003). Do you enjoy having more than others? Survey evidence of positional goods. Working Papers in Economics No. 100, Goteburg University, May. Duesenberry, J. (1949). Income, saving, and the theory of consumer behavior. Cambridge, MA: Harvard University Press. Easterlin, R. (1995). Will raising the incomes of all increase the happiness of all. Journal of Economic Behavior and Organization, 27, 35–47. Frank, R. H. (1984). Are workers paid their marginal products? American Economic Review, 74(September), 549–571. Frank, R. H. (1985). The demand for unobservable and other nonpositional goods. American Economic Review, 75(March), 101–116. Frank, R. H. (1991). Positional externalities. In: R. Zeckhauser (Ed.), Strategy and choice: Essays in honor of Thomas C. Schelling (pp. 25–47). Cambridge, MA: MIT Press. Frank, R. H. (1995). Consumption externalities and the financing of social services. In: V. R. Fuchs (Ed.), Responsible society: Child care, education, medical care, and long-term care in America. NBER. Frank, R. H. (1999). Luxury fever. New York: The Free Press. Frank, R. H., & Levine, A. (2005). Expenditure cascades. Cornell University, Mimeo. Heffetz, O. (2005). Conspicuous consumption and the visibility of consumption expenditures. Department of Economics, Princeton University, Mimeo. Hirsch, F. (1976). Social limits to growth. Cambridge, MA: Harvard University Press. Landers, R., Rebitzer, J., & Taylor, A. (1996). Rat race redux: Adverse selection in the determination of work hours in law firms. American Economic Review, 86, 329–348. Luttmer, E. (2005). Neighbors as negatives: Relative earnings and well-being. Quarterly Journal of Economics, 120(3), 963–1002. Neumark, D., & Postlewaite, A. (1998). Relative income concerns and the rise in married women’ employment. Journal of Public Economics, 70, 157–183. Robson, A. (1992). Status, the distribution of wealth, private and social attitudes to risk. Econometrica, 60, 837–857. Solnick, S., & Hemenway, D. (1998). Is more always better? Journal of Economic Behavior and Organization, 37(3), 373–383. Solnick, S., & Hemenway, D. (2005). Are positional concerns stronger in some domains than in others? American Economic Review, Papers and Proceedings, 95(2), 147–151. Warren, E., & Tyagi, A. W. (2003). The two-income trap: Why middle-class mothers and fathers are going broke. New York: Basic Books.